Yifan Yang, Zheshu Song, Jianheng Zhuo, Mingyu Cui, Jinpeng Li, Bo Yang, Yexing Du, Ziyang Ma, Xunying Liu, Ziyuan Wang, Ke Li, Shuai Fan, Kai Yu, Wei-Qiang Zhang, Guoguo Chen, Xie Chen. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Multi-talker speech recognition (MTASR) faces unique challenges in disentangling and transcribing overlapping speech. To address these challenges, this paper investigates the role of Connectionist Temporal Classification (CTC) in speaker disentanglement when incorporated with Serialized Output Training (SOT) for MTASR. Our visualization reveals that CTC guides the encoder to represent different speakers in distinct temporal regions of acoustic embeddings. Leveraging this insight, we propose a novel Speaker-Aware CTC (SACTC) training objective, based on the Bayes risk CTC framework. SACTC is a tailored CTC variant for multi-talker scenarios, it explicitly models speaker disentanglement by constraining the encoder to represent different speakers’ tokens at specific time frames. When integrated with SOT, the SOT-SACTC model consistently outperforms standard SOT-CTC across various degrees of speech overlap. Specifically, we observe relative word error rate reductions of 10% overall and 15% on low-overlap speech. This work represents an initial exploration of CTC-based enhancements for MTASR tasks, offering a new perspective on speaker disentanglement in multi-talker speech recognition .1

Recent advancements in large language models (LLMs) have revolutionized various domains, bringing significant progress and new opportunities. Despite progress in speech-related tasks, LLMs have not been sufficiently explored in multi-talker scenarios. In this work, we present a pioneering effort to investigate the capability of LLMs in transcribing speech in multi-talker environments, following versatile instructions related to multi-talker automatic speech recognition (ASR), target talker ASR, and ASR based on specific talker attributes such as sex, occurrence order, language, and keyword spoken. Our approach utilizes WavLM and Whisper encoder to extract multi-faceted speech representations that are sensitive to speaker characteristics and semantic context. These representations are then fed into an LLM fine-tuned using LoRA, enabling the capabilities for speech comprehension and transcription. Comprehensive experiments reveal the promising performance of our proposed system, MT-LLM, in cocktail party scenarios, highlighting the potential of LLM to handle speech-related tasks based on user instructions in such complex settings1.

Discrete tokens provide compact and domain-adaptable representations of speech features. However, their application to disordered speech, characterized by articulation imprecision and significant mismatch with normal voice, remains unexplored. To this end, this paper proposes novel phone-purity guided (PPG) discrete tokens to address the weakened phonetic discrimination arising during unsupervised K-means clustering or vector quantization of continuous features. Phonetic label supervision is incorporated to regularize the maximum likelihood and reconstruction error costs in standard K-means and VAE-VQ-based token extraction. Experiments on the UASpeech corpus show that PPG-based discrete tokens extracted from HuBERT consistently outperform hybrid TDNN and End-to-End (E2E) Conformer systems using non-PPG tokens. Statistically significant word error rate (WER) reductions of up to 0.99% and 1.77% absolute (3.21% and 4.82% relative) are achieved across varying codebook sizes for the 16 UASpeech test dysarthric speakers. The lowest WER of 23.25% is obtained by combining systems using complementary token features. Consistent improvements are also observed in phone purity, and t-SNE visualizations demonstrate sharper decision boundaries between K-means/VAE-VQ clusters with the introduction of phone-purity guidance.

This paper presents a novel mixed-precision quantization approach for speech foundation models that tightly integrates mixed-precision learning and quantized model parameter estimation into one single model compression stage. Experiments conducted on LibriSpeech dataset with fine-tuned wav2vec2.0-base and HuBERT-large models suggest the resulting mixed-precision quantized models increased the lossless compression ratio by factors up to 1.7x and 1.9x over the respective uniform-precision and two-stage mixed-precision quantized baselines that perform precision learning and model parameters quantization in separate and disjointed stages, while incurring no statistically word error rate (WER) increase over the 32-bit full-precision models. The system compression time of wav2vec2.0-base and HuBERT-large models is reduced by up to 1.9 and 1.5 times over the two-stage mixed-precision baselines, while both produce lower WERs. The best-performing 3.5-bit mixed-precision quantized HuBERT-large model produces a lossless compression ratio of 8.6x over the 32-bit full-precision system.

Accurate recognition of dysarthric and elderly speech remains challenging to date. While privacy concerns have driven a shift from centralized approaches to federated learning (FL) to ensure data confidentiality, this further exacerbates the challenges of data scarcity, imbalanced data distribution and speaker heterogeneity. To this end, this paper conducts a systematic investigation of regularized FL techniques for privacy-preserving dysarthric and elderly speech recognition, addressing different levels of the FL process by 1) parameter-based, 2) embedding-based and 3) novel loss-based regularization. Experiments on the benchmark UASpeech dysarthric and DementiaBank Pitt elderly speech corpora suggest that regularized FL systems consistently outperform the baseline FedAvg system by statistically significant WER reductions of up to 0.55% absolute (2.13% relative). Further increasing communication frequency to one exchange per batch approaches centralized training performance.

End-to-end multi-talker speech recognition has garnered great interest as an effective approach to directly transcribe overlapped speech from multiple speakers. Current methods typically adopt either 1) single-input multiple-output (SIMO) models with a branched encoder, or 2) single-input single-output (SISO) models based on attention-based encoder-decoder architecture with serialized output training (SOT). In this work, we propose a Cross-Speaker Encoding (CSE) network to address the limitations of SIMO models by aggregating cross-speaker representations. Furthermore, the CSE model is integrated with SOT to leverage both the advantages of SIMO and SISO while mitigating their drawbacks. To the best of our knowledge, this work represents an early effort to integrate SIMO and SISO for multi-talker speech recognition. Experiments on the two-speaker LibrispeechMix dataset show that the CES model reduces word error rate (WER) by 8% over the SIMO baseline. The CSE-SOT model reduces WER by 10% overall and by 16% on high-overlap speech compared to the SOT model.

Dysarthric speech reconstruction (DSR) aims to transform dysarthric speech into normal speech by improving the intelligibility and naturalness. This is a challenging task especially for patients with severe dysarthria and speaking in complex, noisy acoustic environments. To address these challenges, we propose a novel multi-modal framework to utilize visual information, e.g., lip movements, in DSR as extra clues for reconstructing the highly abnormal pronunciations. The multi-modal framework consists of: (i) a multi-modal encoder to extract robust phoneme embeddings from dysarthric speech with auxiliary visual features; (ii) a variance adaptor to infer the normal phoneme duration and pitch contour from the extracted phoneme embeddings; (iii) a speaker encoder to encode the speaker’s voice characteristics; and (iv) a mel-decoder to generate the reconstructed mel-spectrogram based on the extracted phoneme embeddings, prosodic features and speaker embeddings. Both objective and subjective evaluations conducted on the commonly used UASpeech corpus show that our proposed approach can achieve significant improvements over baseline systems in terms of speech intelligibility and naturalness, especially for the speakers with more severe symptoms. Compared with original dysarthric speech, the reconstructed speech achieves 42.1% absolute word error rate reduction for patients with more severe dysarthria levels.1

Practical application of model-based speaker adaptation techniques to end-to-end ASR systems is hindered by speaker-level data scarcity and latency in speaker-dependent (SD) parameters update. To this end, data-efficient and low-latency rapid feature-based speaker adaptation approaches are proposed in this paper for state-of-the-art Conformer ASR systems. Compact subspace projection of training data estimated SD hidden layer output scaling or bias parameters is used to represent the most distinctive speaker "bases". A feature-driven prediction network containing purpose-built speaker-aware memory is designed to on-the-fly produce homogeneous SD basis interpolation, and facilitate rapid speaker adaptation. Experimental results on the 300-hr Switchboard corpus suggest that the proposed adaptation approach produces statistically significant word error rate (WER) reductions of up to 1.0% absolute (8.4% relative) over the baseline speaker-independent and i-vector adapted Conformers before and after external LM rescoring. Consistent WER reductions of up to 2.0% absolute (16.3% relative) and real-time factor speeding up ratios of up to 10.9 times are also obtained over offline model-based adaptation across different speaker-level data quantity operating points. T-SNE visualization reveals the on-the-fly predicted SD basis weights present intuitively more consistent speaker features than i-vectors.

Automatic recognition of disordered speech remains a highly challenging task to date due to data scarcity. This paper presents a reinforcement learning (RL) based on-the-fly data augmentation approach for training state-of-the-art PyChain TDNN and end-to-end Conformer ASR systems on such data. The handcrafted temporal and spectral mask operations in the standard SpecAugment method that are task and system dependent, together with additionally introduced minimum and maximum cut-offs of these masks, are now automatically learned using an RNN-based policy controller and tightly integrated with ASR system training. Experiments on the UASpeech corpus suggest the proposed RL-based data augmentation consistently produced performance superior or comparable to that obtained using expert or handcrafted SpecAugment policies. Our RL auto-augmented PyChain TDNN system produced an overall WER of 28.79% on the UASpeech test set of 16 dysarthric speakers.

Automatic recognition of dysarthric speech remains a highly challenging task to date. Neuro-motor conditions and co-occurring physical disabilities create difficulty in large-scale data collection for ASR system development. Adapting SSL pre-trained ASR models to limited dysarthric speech via data-intensive parameter fine-tuning leads to poor generalization. To this end, this paper presents an extensive comparative study of various data augmentation approaches to improve the robustness of pre-trained ASR model fine-tuning to dysarthric speech. These include: a) conventional speaker-independent perturbation of impaired speech; b) speaker-dependent speed perturbation, or GAN-based adversarial perturbation of normal, control speech based on their time alignment against parallel dysarthric speech; c) novel Spectral basis GAN-based adversarial data augmentation operating on non-parallel data. Experiments conducted on the UASpeech corpus suggest GAN-based data augmentation consistently outperforms fine-tuned Wav2vec2.0 and HuBERT models using no data augmentation and speed perturbation across different data expansion operating points by statistically significant word error rate (WER) reductions up to 2.01% and 0.96% absolute (9.03% and 4.63% relative) respectively on the UASpeech test set of 16 dysarthric speakers. After cross-system outputs rescoring, the best system produced the lowest published WER of 16.53% (46.47% on very low intelligibility) on UASpeech.

Interspeech 2024, 1-5 September 2024, Kos, Greece

We propose a novel one-pass multiple ASR systems joint compression and quantization approach using an all-in-one neural model. A single compression cycle allows multiple nested systems with varying Encoder depths, widths, and quantization precision settings to be simultaneously constructed without the need to train and store individual target systems separately. Experiments consistently demonstrate the multiple ASR systems compressed in a single all-in-one model produced a word error rate (WER) comparable to, or lower by up to 1.01% absolute (6.98% relative) than individually trained systems of equal complexity. A 3.4x overall system compression and training time speed-up was achieved. Maximum model size compression ratios of 12.8x and 3.93x were obtained over the baseline Switchboard-300hr Conformer and LibriSpeech-100hr fine-tuned wav2vec2.0 models, respectively, incurring no statistically significant WER increase.

This paper proposes a novel non-autoregressive (NAR) block-based Attention Mask Decoder (AMD) that flexibly balances performance-efficiency trade-offs for Conformer ASR systems. AMD performs parallel NAR inference within contiguous blocks of output labels that are concealed using attention masks, while conducting left-to-right AR prediction and history context amalgamation between blocks. A beam search algorithm is designed to leverage a dynamic fusion of CTC, AR Decoder, and AMD probabilities. Experiments on the LibriSpeech-100hr corpus suggest the tripartite Decoder incorporating the AMD module produces a maximum decoding speed-up ratio of 1.73x over the baseline CTC+AR decoding, while incurring no statistically significant word error rate (WER) increase on the test sets. When operating with the same decoding real time factors, statistically significant WER reductions of up to 0.7% and 0.3% absolute (5.3% and 6.1% relative) were obtained over the CTC+AR baseline.

Self-supervised learning (SSL) based speech foundation models have been applied to a wide range of ASR tasks. However, their application to dysarthric and elderly speech via data-intensive parameter fine-tuning is confronted by in-domain data scarcity and mismatch. To this end, this paper explores a series of approaches to integrate domain fine-tuned SSL pre-trained models and their features into TDNN and Conformer ASR systems for dysarthric and elderly speech recognition. These include: a) input feature fusion between standard acoustic frontends and domain fine-tuned SSL speech representations; b) frame-level joint decoding between TDNN systems separately trained using standard acoustic features alone and those with additional domain fine-tuned SSL features; and c) multi-pass decoding involving the TDNN/Conformer system outputs to be rescored using domain fine-tuned pre-trained ASR models. In addition, fine-tuned SSL speech features are used in acoustic-to-articulatory (A2A) inversion to construct multi-modal ASR systems. Experiments are conducted on four tasks: the English UASpeech and TORGO dysarthric speech corpora; and the English DementiaBank Pitt and Cantonese JCCOCC MoCA elderly speech datasets. The TDNN systems constructed by integrating domain-adapted HuBERT, wav2vec2-conformer or multi-lingual XLSR models and their features consistently outperform the standalone fine-tuned SSL pre-trained models. These systems produced statistically significant WER or CER reductions of6.53%,1.90%,2.04%and7.97%absolute (24.10%,23.84%,10.14%and31.39%relative) on the four tasks respectively. Consistent improvements in Alzheimer's Disease detection accuracy are also obtained using the DementiaBank Pitt elderly speech recognition outputs.

Despite the rapid progress of automatic speech recognition (ASR) technologies targeting normal speech, accurate recognition of dysarthric and elderly speech remains a highly challenging task to date. It is difficult to collect large quantities of such data for ASR system development due to the mobility issues often found among these users. To this end, data augmentation techniques play a vital role. In contrast to existing data augmentation techniques only modifying the speaking rate or overall shape of spectral contour, fine-grained spectro-temporal differences between dysarthric, elderly and normal speech are modelled using a novel set of speaker-dependent (SD) generative adversarial networks (GAN) based data augmentation approaches in this article. These flexibly allow both: a) temporal or speed perturbed normal speech spectra to be modified and closer to those of an impaired speaker when parallel speech data is available; and b) for non-parallel data, the SVD decomposed normal speech spectral basis features to be transformed into those of a target elderly speaker before being re-composed with the temporal bases to produce the augmented data for state-of-the-art TDNN and Conformer ASR system training. Experiments are conducted on four tasks: the English UASpeech and TORGO dysarthric speech corpora; the English DementiaBank Pitt and Cantonese JCCOCC MoCA elderly speech datasets. The proposed GAN based data augmentation approaches consistently outperform the baseline speed perturbation method by up to 0.91% and 3.0% absolute (9.61% and 6.4% relative) WER reduction on the TORGO and DementiaBank data respectively. Consistent performance improvements are retained after applying LHUC based speaker adaptation.

Automatic recognition of disordered and elderly speech remains a highly challenging task to date due to the difficulty in collecting such data in large quantities. This paper explores a series of approaches to integrate domain adapted Self-Supervised Learning (SSL) pre-trained models into TDNN and Conformer ASR systems for dysarthric and elderly speech recognition: a) input feature fusion between standard acoustic frontends and domain adapted wav2vec2.0 speech representations; b) frame-level joint decoding of TDNN systems separately trained using standard acoustic features alone and with additional wav2vec2.0 features; and c) multi-pass decoding involving the TDNN/Conformer system outputs to be rescored using domain adapted wav2vec2.0 models. In addition, domain adapted wav2vec2.0 representations are utilized in acoustic-to-articulatory (A2A) inversion to construct multi-modal dysarthric and elderly speech recognition systems. Experiments conducted on the UASpeech dysarthric and DementiaBank Pitt elderly speech corpora suggest TDNN and Conformer ASR systems integrated domain adapted wav2vec2.0 models consistently outperform the standalone wav2vec2.0 models by statistically significant WER reductions of 8.22% and 3.43% absolute (26.71% and 15.88% relative) on the two tasks respectively. The lowest published WERs of 22.56% (52.53% on very low intelligibility, 39.09% on unseen words) and 18.17% are obtained on the UASpeech test set of 16 dysarthric speakers, and the DementiaBank Pitt test set respectively.

Automatic recognition of disordered speech remains a highly challenging task to date. The underlying neuro-motor conditions, often compounded with co-occurring physical disabilities, lead to the difficulty in collecting large quantities of impaired speech required for ASR system development. This paper presents novel variational auto-encoder generative adversarial network (VAE-GAN) based personalized disordered speech augmentation approaches that simultaneously learn to encode, generate and discriminate synthesized impaired speech. Separate latent features are derived to learn dysarthric speech characteristics and phoneme context representations. Self-supervised pre-trained Wav2vec 2.0 embedding features are also incorporated. Experiments conducted on the UASpeech corpus suggest the proposed adversarial data augmentation approach consistently outperformed the baseline speed perturbation and non-VAE GAN augmentation methods with trained hybrid TDNN and End-to-end Conformer systems. After LHUC speaker adaptation, the best system using VAE-GAN based augmentation produced an overall WER of 27.78% on the UASpeech test set of 16 dysarthric speakers, and the lowest published WER of 57.31% on the subset of speakers with "Very Low" intelligibility.

With the global population aging rapidly, Alzheimer’s disease (AD) is particularly prominent in older adults, which has an insidious onset and leads to a gradual, irreversible deterioration in cognitive domains (memory, communication, etc.). Speech-based AD detection opens up the possibility of widespread screening and timely disease intervention. Recent advances in pre-trained models motivate AD detection modeling to shift from low-level features to high-level representations. This paper presents several efficient methods to extract better AD-related cues from high-level acoustic and linguistic features. Based on these features, the paper also proposes a novel task-oriented approach by modeling the relationship between the participants’ description and the cognitive task. Experiments are carried out on the ADReSS dataset in a binary classification setup, and models are evaluated on the unseen test set. Results and comparison with recent literature demonstrate the efficiency and superior performance of proposed acoustic, linguistic and task-oriented methods. The findings also show the importance of semantic and syntactic information, and feasibility of automation and generalization with the promising audio-only and task-oriented methods for the AD detection task.

As a common way of emotion signaling via non-linguistic vocalizations, vocal burst (VB) plays an important role in daily social interaction. Understanding and modeling human vocal bursts are indispensable for developing robust and general artificial intelligence. Exploring computational approaches for understanding vocal bursts is attracting increasing research attention. In this work, we propose a hierarchical framework, based on chain regression models, for affective recognition from VBs, that explicitly considers multiple relationships: (i) between emotional states and diverse cultures; (ii) between low-dimensional (arousal & valence) and high-dimensional (10 emotion classes) emotion spaces; and (iii) between various emotion classes within the high-dimensional space. To address the challenge of data sparsity, we also use self-supervised learning (SSL) representations with layer-wise and temporal aggregation modules. The proposed systems participated in the ACII Affective Vocal Burst (A-VB) Challenge 2022 and ranked first in the "TWO" and "CULTURE" tasks. Experimental results based on the ACII Challenge 2022 dataset demonstrate the superior performance of the proposed system and the effectiveness of considering multiple relationships using hierarchical regression chain models.

Early diagnosis of Alzheimer’s disease (AD) is crucial in facilitating preventive care and to delay further progression. Speech based automatic AD screening systems provide a non-intrusive and more scalable alternative to other clinical screening techniques. Textual embedding features produced by pre-trained language models (PLMs) such as BERT are widely used in such systems. However, PLM domain fine-tuning is commonly based on the masked word or sentence prediction costs that are inconsistent with the back-end AD detection task. To this end, this paper investigates the use of prompt-based fine-tuning of PLMs that consistently uses AD classification errors as the training objective function. Disfluency features based on hesitation or pause filler token frequencies are further incorporated into prompt phrases during PLM fine-tuning. The decision voting based combination among systems using different PLMs (BERT and RoBERTa) or systems with different fine-tuning paradigms (conventional masked-language modelling fine-tuning and prompt-based fine-tuning) is further applied. Mean, standard deviation (std) and the maximum among accuracy scores over 15 experiment runs are adopted as performance measurements for the AD detection system. Mean detection accuracy of 84.20% (with std 2.09%, best 87.5%) and 82.64% (with std 4.0%, best 89.58%) were obtained using manual and ASR speech transcripts respectively on the ADReSS20 test set consisting of 48 elderly speakers.

Modeling the speaker variability is a key challenge for automatic speech recognition (ASR) systems. In this paper, the learning hidden unit contributions (LHUC) based adaptation techniques with compact speaker dependent (SD) parameters are used to facilitate both speaker adaptive training (SAT) and unsupervised test-time speaker adaptation for end-to-end (E2E) lattice-free MMI (LF-MMI) models. An unsupervised model-based adaptation framework is proposed to estimate the SD parameters in E2E paradigm using LF-MMI and cross entropy (CE) criterions. Various regularization methods of the standard LHUC adaptation, e.g., the Bayesian LHUC (BLHUC) adaptation, are systematically investigated to mitigate the risk of overfitting, on E2E LF-MMI CNN-TDNN and CNN-TDNN- BLSTM models. Lattice-based confidence score estimation is used for adaptation data selection to reduce the supervision label uncertainty. Experiments on the 300-hour Switchboard task suggest that, applying BLHUC in the proposed unsupervised E2E adaptation framework to byte pair encoding (BPE) based E2E LF-MMI systems consistently outperformed the baseline systems by relative word error rate (WER) reductions up to 10.5% and 14.7% on the NIST Hub5’00 and RT03 evaluation sets, and achieved the best performance in WERs of 9.0% and 9.7%, respectively. These results are comparable to the results of state-of-the-art adapted LF-MMI hybrid systems and adapted Conformer-based E2E systems.

24th Annual Conference of the International Speech Communication Association, INTERSPEECH 2023, Dublin, Ireland, August 20-24, 2023

Speaker adaptation techniques provide a powerful solution to customise automatic speech recognition (ASR) systems for individual users. Practical application of unsupervised model-based speaker adaptation techniques to data intensive end-to-end ASR systems is hindered by the scarcity of speaker-level data and performance sensitivity to transcription errors. To address these issues, a set of compact and data efficient speaker-dependent (SD) parameter representations are used to facilitate both speaker adaptive training and test-time unsupervised speaker adaptation of state-of-the-art Conformer ASR systems. The sensitivity to supervision quality is reduced using a confidence score-based selection of the less erroneous subset of speaker-level adaptation data. Two lightweight confidence score estimation modules are proposed to produce more reliable confidence scores. The data sparsity issue, which is exacerbated by data selection, is addressed by modelling the SD parameter uncertainty using Bayesian learning. Experiments on the benchmark 300-hour Switchboard and the 233-hour AMI datasets suggest that the proposed confidence score-based adaptation schemes consistently outperformed the baseline speaker-independent (SI) Conformer model and conventional non-Bayesian, point estimate-based adaptation using no speaker data selection. Similar consistent performance improvements were retained after external Transformer and LSTM language model rescoring. In particular, on the 300-hour Switchboard corpus, statistically significant WER reductions of 1.0%, 1.3%, and 1.4% absolute (9.5%, 10.9%, and 11.3% relative) were obtained over the baseline SI Conformer on the NIST Hub5’00, RT02, and RT03 evaluation sets respectively. Similar WER reductions of 2.7% and 3.3% absolute (8.9% and 10.2% relative) were also obtained on the AMI development and evaluation sets.

Accurate recognition of cocktail party speech containing overlapping speakers, noise and reverberation remains a highly challenging task to date. Motivated by the invariance of visual modality to acoustic signal corruption, an audio-visual multi-channel speech separation, dereverberation and recognition approach featuring a full incorporation of visual information into all system components is proposed in this paper. The efficacy of the video input is consistently demonstrated in mask-based MVDR speech separation, DNN-WPE or spectral mapping (SpecM) based speech dereverberation front-end and Conformer ASR back-end. Audio-visual integrated front-end architectures performing speech separation and dereverberation in a pipelined or joint fashion via mask-based WPD are investigated. The error cost mismatch between the speech enhancement front-end and ASR back-end components is minimized by end-to-end jointly fine-tuning using either the ASR cost function alone, or its interpolation with the speech enhancement loss. Experiments were conducted on the mixture overlapped and reverberant speech data constructed using simulation or replay of the Oxford LRS2 dataset. The proposed audio-visual multi-channel speech separation, dereverberation and recognition systems consistently outperformed the comparable audio-only baseline by 9.1% and 6.2% absolute (41.7% and 36.0% relative) word error rate (WER) reductions. Consistent speech enhancement improvements were also obtained on PESQ, STOI and SRMR scores.

The attention-based encoder-decoder structure, such as the Transformer, has achieved state-of-the-art performance on various sequence modeling tasks,e.g., machine translation (MT) and automatic speech recognition (ASR), benefited from the superior capability of layer-wise self-attention mechanism in the encoder/decoder to access long-distance contextual information. Recently, analysis on the Transformer layers has shown that different levels of information,e.g., phoneme level, word level and semantic level, are represented at different layers. Effectively integrating information from various levels is important for structured prediction. However, the self-attention in the conventional Transformer structure only focuses on intra-layer integration, and does not explicitly model inter-layer information relationships. Also, attention across the encoder and decoder (cross-coder) only focuses on the top encoder layer but ignores the intermediate layers. In this paper, we propose a sequence modeling structure equipped with a hierarchical attention mechanism, named Hiformer, that can consider the inter-layer and cross-coder hierarchical information to improve structured prediction performance. Extensive experiments conducted on both MT and ASR tasks demonstrate the effectiveness of the proposed Hiformer model.

Articulatory features are inherently invariant to acoustic signal distortion and have been successfully incorporated into automatic speech recognition (ASR) systems for normal speech. Their practical application to disordered speech recognition is often limited by the difficulty in collecting such specialist data from impaired speakers. This paper presents a cross-domain acoustic-to-articulatory (A2A) inversion approach that utilizes the parallel acoustic-articulatory data of the 15-hour TORGO corpus in model training before being cross-domain adapted to the 102.7-hour UASpeech corpus and to produce articulatory features. Mixture density networks based neural A2A inversion models were used. A cross-domain feature adaptation network was also used to reduce the acoustic mismatch between the TORGO and UASpeech data. On both tasks, incorporating the A2A generated articulatory features consistently outperformed the baseline hybrid DNN/TDNN, CTC and Conformer based end-to-end systems constructed using acoustic features only. The best multi-modal system incorporating video modality and the cross-domain articulatory features as well as data augmentation and learning hidden unit contributions (LHUC) speaker adaptation produced the lowest published word error rate (WER) of 24.82% on the 16 dysarthric speakers of the benchmark UASpeech task.

Despite the rapid advance of automatic speech recognition (ASR) technologies, accurate recognition of cocktail party speech characterised by the interference from overlapping speakers, background noise and room reverberation remains a highly challenging task to date. Motivated by the invariance of visual modality to acoustic signal corruption, audio-visual speech enhancement techniques have been developed, although predominantly targeting overlapping speech separation and recognition tasks. In this paper, an audio-visual multi-channel speech separation, dereverberation and recognition approach featuring a full incorporation of visual information into all three stages of the system is proposed. The advantage of the additional visual modality over using audio only is demonstrated on two neural dereverberation approaches based on DNN-WPE and spectral mapping respectively. The learning cost function mismatch between the separation and dereverberation models and their integration with the back-end recognition system is minimised using fine-tuning on the MSE and LF-MMI criteria. Experiments conducted on the LRS2 dataset suggest that the proposed audio-visual multi-channel speech separation, dereverberation and recognition system outperforms the baseline audio-visual multi-channel speech separation and recognition system containing no dereverberation module by a statistically significant word error rate (WER) reduction of 2.06 % absolute (8.77 % relative).

Speaker Change Detection (SCD) is a task of determining the time boundaries between speech segments of different speakers. SCD system can be applied to many tasks, such as speaker diarization, speaker tracking, and transcribing audio with multiple speakers. Recent advancements in deep learning lead to approaches that can directly detect the speaker change points from audio data at the frame-level based on neural network models. These approaches may be further improved by utilizing speaker information in the training data, and utilizing content information extracted in an unsupervised manner. This work proposes a novel framework for the SCD task, which utilizes a multitask learning architecture to leverage speaker information during the training stage, and adds the content information extracted from an unsupervised speech decomposition model to help detect the speaker change points. Experiment results show that the architecture of multitask learning with speaker information can improve the performance of SCD, and adding content information extracted from unsupervised speech decomposition model can further improve the performance. To the best of our knowledge, this work outperforms the state-of-the-art SCD results [1] on the AMI dataset.

Dysarthric speech reconstruction (DSR), which aims to improve the quality of dysarthric speech, remains a challenge, not only because we need to restore the speech to be normal, but also must preserve the speaker’s identity. The speaker representation extracted by the speaker encoder (SE) optimized for speaker verification has been explored to control the speaker identity. However, the SE may not be able to fully capture the characteristics of dysarthric speakers that are previously unseen. To address this research problem, we propose a novel multi-task learning strategy, i.e., adversarial speaker adaptation (ASA). The primary task of ASA fine-tunes the SE with the speech of the target dysarthric speaker to effectively capture identityrelated information, and the secondary task applies adversarial training to avoid the incorporation of abnormal speaking patterns into the reconstructed speech, by regularizing the distribution of reconstructed speech to be close to that of reference speech with high quality. Experiments show that the proposed approach can achieve enhanced speaker similarity and comparable speech naturalness with a strong baseline approach. Compared with dysarthric speech, the reconstructed speech achieves 22.3% and 31.5% absolute word error rate reduction for speakers with moderate and moderate-severe dysarthria respectively. Our demo page is released here1.

Though significant progress has been made for speaker-dependent Video-to-Speech (VTS) synthesis, little attention is devoted to multi-speaker VTS that can map silent video to speech, while allowing flexible control of speaker identity, all in a single system. This paper proposes a novel multi-speaker VTS system based on cross-modal knowledge transfer from voice conversion (VC), where vector quantization with contrastive predictive coding (VQCPC) is used for the content encoder of VC to derive discrete phoneme-like acoustic units, which are transferred to a Lip-to-Index (Lip2Ind) network to infer the index sequence of acoustic units. The Lip2Ind network can then substitute the content encoder of VC to form a multi-speaker VTS system to convert silent video to acoustic units for reconstructing accurate spoken content. The VTS system also inherits the advantages of VC by using a speaker encoder to produce speaker representations to effectively control the speaker identity of generated speech. Extensive evaluations verify the effectiveness of proposed approach, which can be applied in both constrained vocabulary and open vocabulary conditions, achieving state-of-the-art performance in generating high-quality speech with high naturalness, intelligibility and speaker similarity. Our demo page is released here1.

Deep neural networks have brought significant advancements to speech emotion recognition (SER). However, the architecture design in SER is mainly based on expert knowledge and empirical (trial-and-error) evaluations, which is time-consuming and resource intensive. In this paper, we propose to apply neural architecture search (NAS) techniques to automatically configure the SER models. To accelerate the candidate architecture optimization, we propose a uniform path dropout strategy to encourage all candidate architecture operations to be equally optimized. Experimental results of two different neural structures on IEMOCAP show that NAS can improve SER performance (54.89% to 56.28%) while maintaining model parameter sizes. The proposed dropout strategy also shows superiority over the previous approaches.

Recognition of overlapped speech has been a highly challenging task to date. State-of-the-art multi-channel speech separation system are becoming increasingly complex and expensive for practical applications. To this end, low-bit neural network quantization provides a powerful solution to dramatically reduce their model size. However, current quantization methods are based on uniform precision and fail to account for the varying performance sensitivity at different model components to quantization errors. In this paper, novel mixed precision DNN quantization methods are proposed by applying locally variable bit-widths to individual TCN components of a TF masking based multi-channel speech separation system. The optimal local precision settings are automatically learned using three techniques. The first two approaches utilize quantization sensitivity metrics based on either the mean square error (MSE) loss function curvature, or the KL-divergence measured between full precision and quantized separation models. The third approach is based on mixed precision neural architecture search. Experiments conducted on the LRS3-TED corpus simulated overlapped speech data suggest that the proposed mixed precision quantization techniques consistently outperform the uniform precision baseline speech separation systems of comparable bit-widths in terms of SI-SNR and PESQ scores as well as word error rate (WER) reductions up to 2.88% absolute (8% relative).

Speaker identification for overlapped speech presents a great challenge for speaker diarization tasks in meeting scenarios. In order to overcome such challenges, several overlap-aware resegmentation methods based on deep learning have been integrated into speaker diarization systems. In this paper we propose two multi-channel diarization systems which have enhanced capability in detecting overlapped speech and identify speakers via learning spatial features. The first system applies a multi-look strategy to train networks without given the speakers’ direction of arrival(DOA), and the other system estimates the DOA of target speakers based on existing diarization results. Both systems aim to estimate the voice activity of speakers in different directions to handle overlapped speech. Experimental results on the AMI corpus show that the relative improvements of both systems can reach 9.4% and 18.1% in term of diarization error rate (DER) against an overlap-aware single-channel system with a BeamformIt front-end.

Fundamental modelling differences between hybrid and end-to-end (E2E) automatic speech recognition (ASR) systems create large diversity and complementarity among them. This paper investigates multi-pass rescoring and cross adaptation based system combination approaches for hybrid TDNN and Conformer E2E ASR systems. In multi-pass rescoring, state-of-the-art hybrid LF-MMI trained CNN-TDNN system featuring speed perturbation, SpecAugment and Bayesian learning hidden unit contributions (LHUC) speaker adaptation was used to produce initial N-best outputs before being rescored by the speaker adapted Conformer system using a 2-way cross system score interpolation. In cross adaptation, the hybrid CNN-TDNN system was adapted to the 1-best output of the Conformer system or vice versa. Experiments on the 300-hour Switchboard corpus suggest that the combined systems derived using either of the two system combination approaches outperformed the individual systems. The best combined system obtained using multi-pass rescoring produced statistically significant word error rate (WER) reductions of 2.5% to 3.9% absolute (22.5% to 28.9% relative) over the stand alone Conformer system on the NIST Hub5'00, Rt03 and Rt02 evaluation data.

A key challenge for automatic speech recognition (ASR) systems is to model the speaker level variability.In this paper, compact speaker dependent learning hidden unit contributions (LHUC) are used to facilitate both speaker adaptive training (SAT) and test time unsupervised speaker adaptation for stateof-the-art Conformer based end-to-end ASR systems.The sensitivity during adaptation to supervision error rate is reduced using confidence score based selection of the more "trustworthy" subset of speaker specific data.A confidence estimation module is used to smooth the over-confident Conformer decoder output probabilities before serving as confidence scores.The increased data sparsity due to speaker level data selection is addressed using Bayesian estimation of LHUC parameters.Experiments on the 300-hour Switchboard corpus suggest that the proposed LHUC-SAT Conformer with confidence score based test time unsupervised adaptation outperformed the baseline speaker independent and i-vector adapted Conformer systems by up to 1.0%, 1.0%, and 1.2% absolute (9.0%, 7.9%, and 8.9% relative) word error rate (WER) reductions on the NIST Hub5'00, RT02, and RT03 evaluation sets respectively.Consistent performance improvements were retained after external Transformer and LSTM language models were used for rescoring.

Early diagnosis of Alzheimer's disease (AD) is crucial in facilitating preventive care to delay further progression. This paper presents the development of a state-of-the-art Conformer based speech recognition system built on the DementiaBank Pitt corpus for automatic AD detection. The baseline Conformer system trained with speed perturbation and SpecAugment based data augmentation is significantly improved by incorporating a set of purposefully designed modeling features, including neural architecture search based auto-configuration of domain-specific Conformer hyper-parameters in addition to parameter fine-tuning; fine-grained elderly speaker adaptation using learning hidden unit contributions (LHUC); and two-pass cross-system rescoring based combination with hybrid TDNN systems. An overall word error rate (WER) reduction of 13.6% absolute (34.8% relative) was obtained on the evaluation data of 48 elderly speakers. Using the final systems' recognition outputs to extract textual features, the best-published speech recognition based AD detection accuracy of 91.7% was obtained.

Early diagnosis of Alzheimer's disease (AD) is crucial in facilitating preventive care and delay progression. Speech based automatic AD screening systems provide a non-intrusive and more scalable alternative to other clinical screening techniques. Scarcity of such specialist data leads to uncertainty in both model selection and feature learning when developing such systems. To this end, this paper investigates the use of feature and model combination approaches to improve the robustness of domain fine-tuning of BERT and Roberta pre-trained text encoders on limited data, before the resulting embedding features being fed into an ensemble of backend classifiers to produce the final AD detection decision via majority voting. Experiments conducted on the ADReSS20 Challenge dataset suggest consistent performance improvements were obtained using model and feature combination in system development. State-of-the-art AD detection accuracies of 91.67 percent and 93.75 percent were obtained using manual and ASR speech transcripts respectively on the ADReSS20 test set consisting of 48 elderly speakers.

We propose an unsupervised learning method to disentangle speech into content representation and speaker identity representation. We apply this method to the challenging one-shot cross-lingual voice conversion task to demonstrate the effectiveness of the disentanglement. Inspired by ß- VAE, we introduce a learning objective that balances between the information captured by the content and speaker representations. In addition, the inductive biases from the architectural design and the training dataset further encourage the desired disentanglement. Both objective and subjective evaluations show the effectiveness of the proposed method in speech disentanglement and in one-shot cross-lingual voice conversion.

Despite the rapid progress of automatic speech recognition (ASR) technologies targeting normal speech in recent decades, accurate recognition of dysarthric and elderly speech remains highly challenging tasks to date. Sources of heterogeneity commonly found in normal speech including accent or gender, when further compounded with the variability over age and speech pathology severity level, create large diversity among speakers. To this end, speaker adaptation techniques play a key role in personalization of ASR systems for such users. Motivated by the spectro-temporal level differences between dysarthric, elderly and normal speech that systematically manifest in articulatory imprecision, decreased volume and clarity, slower speaking rates and increased dysfluencies, novel spectro-temporal subspace basis deep embedding features derived using SVD speech spectrum decomposition are proposed in this paper to facilitate auxiliary feature based speaker adaptation of state-of-the-art hybrid DNN/TDNN and end-to-end Conformer speech recognition systems. Experiments were conducted on four tasks: the English UASpeech and TORGO dysarthric speech corpora; the English DementiaBank Pitt and Cantonese JCCOCC MoCA elderly speech datasets. The proposed spectro-temporal deep feature adapted systems outperformed baseline i-Vector and x-Vector adaptation by up to 2.63% absolute (8.63% relative) reduction in word error rate (WER). Consistent performance improvements were retained after model based speaker adaptation using learning hidden unit contributions (LHUC) was further applied. The best speaker adapted system using the proposed spectral basis embedding features produced the lowest published WER of 25.05% on the UASpeech test set of 16 dysarthric speakers.

State-of-the-art automatic speech recognition (ASR) system development is data and computation intensive. The optimal design of deep neural networks (DNNs) for these systems often require expert knowledge and empirical evaluation. In this paper, a range of neural architecture search (NAS) techniques are used to automatically learn two types of hyper-parameters of factored time delay neural networks (TDNN-Fs): i) the left and right splicing context offsets; and ii) the dimensionality of the bottleneck linear projection at each hidden layer. These techniques include the differentiable neural architecture search (DARTS) method integrating architecture learning with lattice-free MMI training; Gumbel-Softmax and pipelined DARTS methods reducing the confusion over candidate architectures and improving the generalization of architecture selection; and Penalized DARTS incorporating resource constraints to balance the trade-off between performance and system complexity. Parameter sharing among TDNN-F architectures allows an efficient search over up to$7^{28}$different systems. Statistically significant word error rate (WER) reductions of up to 1.2% absolute and relative model size reduction of 31% were obtained over a state-of-the-art 300-hour Switchboard corpus trained baseline LF-MMI TDNN-F system featuring speed perturbation, i-Vector and learning hidden unit contribution (LHUC) based speaker adaptation as well as RNNLM rescoring. Performance contrasts on the same task against recent end-to-end systems reported in the literature suggest the best NAS auto-configured system achieves state-of-the-art WERs of 9.9% and 11.1% on the NIST Hub5’ 00 and Rt03 s test sets respectively with up to 96% model size reduction. Further analysis using Bayesian learning shows that the proposed NAS approaches can effectively minimize the structural redundancy in the TDNN-F systems and reduce their model parameter uncertainty. Consistent performance improvements were also obtained on a UASpeech dysarthric speech recognition task.

State-of-the-art neural network language models (NNLMs) represented by long short term memory recurrent neural networks (LSTM-RNNs) and Transformers are becoming highly complex. They are prone to overfitting and poor generalization when given limited training data. To this end, an overarching full Bayesian learning framework encompassing three methods is proposed in this paper to account for the underlying uncertainty in LSTM-RNN and Transformer LMs. The uncertainty over their model parameters, choice of neural activations and hidden output representations are modeled using Bayesian, Gaussian Process and variational LSTM-RNN or Transformer LMs respectively. Efficient inference approaches were used to automatically select the optimal network internal components to be Bayesian learned using neural architecture search. A minimal number of Monte Carlo parameter samples as low as one was also used. These allow the computational costs incurred in Bayesian NNLM training and evaluation to be minimized. Experiments are conducted on two tasks: AMI meeting transcription and Oxford-BBC LipReading Sentences 2 (LRS2) overlapped speech recognition using state-of-the-art LF-MMI trained factored TDNN systems featuring data augmentation, speaker adaptation and audio-visual multi-channel beamforming for overlapped speech. Consistent performance improvements over the baseline LSTM-RNN and Transformer LMs with point estimated model parameters and drop-out regularization were obtained across both tasks in terms of perplexity and word error rate (WER). In particular, on the LRS2 data, statistically significant WER reductions up to 1.3% and 1.2% absolute (12.1% and 11.3% relative) were obtained over the baseline LSTM-RNN and Transformer LMs respectively after model combination between Bayesian NNLMs and their respective baselines.

Controversy exists on whether differentiable neural architecture search methods discover wiring topology effectively. To understand how wiring topology evolves, we study the underlying mechanism of several existing differentiable NAS frameworks. Our investigation is motivated by three observed searching patterns of differentiable NAS: 1) they search by growing instead of pruning; 2) wider networks are more preferred than deeper ones; 3) no edges are selected in bi-level optimization. To anatomize these phenomena, we propose a unified view on searching algorithms of existing frameworks, transferring the global optimization to local cost minimization. Based on this reformulation, we conduct empirical and theoretical analyses, revealing implicit biases in the cost’s assignment mechanism and evolution dynamics that cause the observed phenomena. These biases indicate strong discrimination towards certain topologies. To this end, we pose questions that future differentiable methods for neural wiring discovery need to confront, hoping to evoke a discussion and rethinking on how much bias has been enforced implicitly in existing NAS methods.

Deep neural networks (DNNs) based automatic speech recognition (ASR) systems are often designed using expert knowledge and empirical evaluation. In this paper, a range of neural architecture search (NAS) techniques are used to automatically learn two types of hyper-parameters of state-of-the-art factored time delay neural networks (TDNNs): i) the left and right splicing context offsets; and ii) the dimensionality of the bottleneck linear projection at each hidden layer. These include the DARTS method integrating architecture selection with lattice-free MMI (LF-MMI) TDNN training; Gumbel-Softmax and pipelined DARTS reducing the confusion over candidate architectures and improving the generalization of architecture selection; and Penalized DARTS incorporating resource constraints to adjust the trade-off between performance and system complexity. Parameter sharing among candidate architectures allows efficient search over up to 728different TDNN systems. Experiments conducted on the 300-hour Switchboard corpus suggest the auto-configured systems consistently outperform the baseline LF-MMI TDNN systems using manual network design or random architecture search after LHUC speaker adaptation and RNNLM rescoring. Absolute word error rate (WER) reductions up to 1.0% and relative model size reduction of 28% were obtained. Consistent performance improvements were also obtained on a UASpeech disordered speech recognition task using the proposed NAS approaches.

Existing approaches for replay and synthetic speech detection still lack generalizability to unseen spoofing attacks. This work proposes to leverage a novel model structure, so-called Res2Net, to improve the anti-spoofing countermeasure’s generalizability. Res2Net mainly modifies the ResNet block to enable multiple feature scales. Specifically, it splits the feature maps within one block into multiple channel groups and designs a residual-like connection across different channel groups. Such connection increases the possible receptive fields, resulting in multiple feature scales. This multiple scaling mechanism significantly improves the countermeasure’s generalizability to unseen spoofing attacks. It also decreases the model size compared to ResNet-based models. Experimental results show that the Res2Net model consistently outperforms ResNet34 and ResNet50 by a large margin in both physical access (PA) and logical access (LA) of the ASVspoof 2019 corpus. Moreover, integration with the squeeze-and-excitation (SE) block can further enhance performance. For feature engineering, we investigate the gen-eralizability of Res2Net combined with different acoustic features, and observe that the constant-Q transform (CQT) achieves the most promising performance in both PA and LA scenarios. Our best single system outperforms other state-of-the-art single systems in both PA and LA of the ASVspoof 2019 corpus.

With the global population ageing rapidly, Alzheimer's disease (AD) is particularly prominent in older adults, which has an insidious onset followed by gradual, irreversible deterioration in cognitive domains (memory, communication, etc). Thus the detection of Alzheimer's disease is crucial for timely intervention to slow down disease progression. This paper presents a comparative study of different acoustic and linguistic features for the AD detection using various classifiers. Experimental results on ADReSS dataset reflect that the proposed models using ComParE, X-vector, Linguistics, TFIDF and BERT features are able to detect AD with high accuracy and sensitivity, and are comparable with the state-of-the-art results reported. While most previous work used manual transcripts, our results also indicate that similar or even better performance could be obtained using automatically recognized transcripts over manually collected ones. This work achieves accuracy scores at 0.67 for acoustic features and 0.88 for linguistic features on either manual or ASR transcripts on the ADReSS Challenge1test set.

Sequence-to-sequence (seq2seq) learning has greatly improved text-to-speech (TTS) synthesis performance, but effective implementation on resource-restricted devices remains challenging as seq2seq models are usually computationally expensive and memory intensive. To achieve fast inference speed and small model size while maintain high-quality speech, we propose FCL-taco2, a Fast, Controllable and Lightweight (FCL) TTS model based on Tacotron2. FCL-taco2 adopts a novel semi-autoregressive (SAR) mode for phoneme level based parallel mel-spectrograms generation conditioned on prosody features, leading to faster inference speed and higher prosody controllability than Tacotron2. Besides, knowledge distillation (KD) is leveraged to compress a relatively large FCL-taco2 model to its small version with minor loss of speech quality. Experimental results on English (EN) and Chinese (CN) datasets show that the small version of FCL-taco2 achieves comparable performance with Tacotron2 in terms of speech quality, while it has a 4.8× smaller footprint with 17.7× and 18.5× faster inference speeds on average for EN and CN experiments respectively. Besides, execution on mobile devices shows that the proposed model can achieve faster than real-time speech synthesis. Our code and audio samples are released1.

State-of-the-art neural language models represented by Transformers are becoming increasingly complex and expensive for practical applications. Low-bit deep neural network quantization techniques provides a powerful solution to dramatically reduce their model size. Current low-bit quantization methods are based on uniform precision and fail to account for the varying performance sensitivity at different parts of the system to quantization errors. To this end, novel mixed precision DNN quantization methods are proposed in this paper. The optimal local precision settings are automatically learned using two techniques. The first is based on a quantization sensitivity metric in the form of Hessian trace weighted quantization perturbation. The second is based on mixed precision Transformer architecture search. Alternating direction methods of multipliers (ADMM) are used to efficiently train mixed precision quantized DNN systems. Experiments conducted on Penn Treebank (PTB) and a Switchboard corpus trained LF-MMI TDNN system suggest the proposed mixed precision Transformer quantization techniques achieved model size compression ratios of up to 16 times over the full precision baseline with no recognition performance degradation. When being used to compress a larger full precision Transformer LM with more layers, overall word error rate (WER) reductions up to 1.7% absolute (18% relative) were obtained.

State-of-the-art neural language models (LMs) represented by Transformers are highly complex. Their use of fixed, deterministic parameter estimates fail to account for model uncertainty and lead to over-fitting and poor generalization when given limited training data. In order to address these issues, this paper proposes a full Bayesian learning framework for Transformer LM estimation. Efficient variational inference based approaches are used to estimate the latent parameter posterior distributions associated with different parts of the Transformer model architecture including multi-head self-attention, feed forward and embedding layers. Statistically significant word error rate (WER) reductions up to 0.5% absolute (3.18% relative) and consistent perplexity gains were obtained over the baseline Transformer LMs on state-of-the-art Switchboard corpus trained LF-MMI factored TDNN systems with i-Vector speaker adaptation. Performance improvements were also obtained on a cross domain LM adaptation task requiring porting a Transformer LM trained on the Switchboard and Fisher data to a low-resource DementiaBank elderly speech corpus.

Early diagnosis of Neurocognitive Disorder (NCD) is crucial in facilitating preventive care and timely treatment to delay further progression. This paper presents the development of a state-of-the-art automatic speech recognition (ASR) system built on the Dementia-Bank Pitt corpus for automatic NCD detection. Speed perturbation based audio data augmentation expanded the limited elderly speech data by four times. Large quantities of out-of-domain, non-aged adult speech were exploited by cross-domain adapting a 1000-hour LibriSpeech corpus trained LF-MMI factored TDNN system to DementiaBank. The variability among elderly speakers was modeled using i-Vector and learning hidden unit contributions (LHUC) based speaker adaptive training. Robust Bayesian estimation of TDNN systems and LHUC transforms were used in both cross-domain and speaker adaptation. A Transformer language model was also built to improve the final system performance. A word error rate (WER) reduction of 11.72% absolute (26.11% relative) was obtained over the baseline i-Vector adapted LF-MMI TDNN system on the evaluation data of 48 elderly speakers. The best NCD detection accuracy of 88%, comparable to that using the ground truth speech transcripts, was obtained using the textual features extracted from the final ASR system outputs.

In multi-talker cases, overlapped speech degrades the speaker verification (SV) performance dramatically. To tackle this challenging problem, speech separation with multi-channel techniques can be adopted to extract each speaker’s signals to improve the SV performance. In this paper, a joint training framework of the front-end multi-look speech separator and the back-end speaker embedding extractor is proposed for multi-channel overlapped speech. To better leverage the complementarity between the speech separator and the speaker embedding extractor, several training strategies are proposed to jointly optimize the two modules. Experimental results show that the proposed joint training framework significantly outperforms the individual SV system by around 52% relative EER reduction. Additionally, the robustness of the proposed framework is further evaluated under different conditions.

Automatic recognition of disordered speech remains a highly challenging task to date. Sources of variability commonly found in normal speech including accent, age or gender, when further compounded with the underlying causes of speech impairment and varying severity levels, create large diversity among speakers. To this end, speaker adaptation techniques play a vital role in current speech recognition systems. Motivated by the spectro-temporal level differences between disordered and normal speech that systematically manifest in articulatory imprecision, decreased volume and clarity, slower speaking rates and increased dysfluencies, novel spectro-temporal subspace basis embedding deep features derived by SVD decomposition of speech spectrum are proposed to facilitate both accurate speech intelligibility assessment and auxiliary feature based speaker adaptation of state-of-the-art hybrid DNN and end-to-end disordered speech recognition systems. Experiments conducted on the UASpeech corpus suggest the proposed spectro-temporal deep feature adapted systems consistently outperformed baseline i-Vector adaptation by up to 2.63% absolute (8.6% relative) reduction in word error rate (WER) with or without data augmentation. Learning hidden unit contribution (LHUC) based speaker adaptation was further applied. The final speaker adapted system using the proposed spectral basis embedding features gave an overall WER of 25.6% on the UASpeech test set of 16 dysarthric speakers

Automatic recognition of disordered speech remains a highly challenging task to date. The underlying neuro-motor conditions, often compounded with co-occurring physical disabilities, lead to the difficulty in collecting large quantities of impaired speech required for ASR system development. To this end, data augmentation techniques play a vital role in current disordered speech recognition systems. In contrast to existing data augmentation techniques only modifying the speaking rate or overall shape of spectral contour, fine-grained spectro-temporal differences between disordered and normal speech are modelled using deep convolutional generative adversarial networks (DCGAN) during data augmentation to modify normal speech spectra into those closer to disordered speech. Experiments conducted on the UASpeech corpus suggest the proposed adversarial data augmentation approach consistently outperformed the baseline augmentation methods using tempo or speed perturbation on a state-of-the-art hybrid DNN system. An overall word error rate (WER) reduction up to 3.05\% (9.7\% relative) was obtained over the baseline system using no data augmentation. The final learning hidden unit contribution (LHUC) speaker adapted system using the best adversarial augmentation approach gives an overall WER of 25.89% on the UASpeech test set of 16 dysarthric speakers.

Existing approaches for anti-spoofing in automatic speaker verification (ASV) still lack generalizability to unseen attacks.The Res2Net approach designs a residual-like connection between feature groups within one block, which increases the possible receptive fields and improves the system's detection generalizability.However, such a residual-like connection is performed by a direct addition between feature groups without channelwise priority.We argue that the information across channels may not contribute to spoofing cues equally, and the less relevant channels are expected to be suppressed before adding onto the next feature group, so that the system can generalize better to unseen attacks.This argument motivates the current work that presents a novel, channel-wise gated Res2Net (CG-Res2Net), which modifies Res2Net to enable a channel-wise gating mechanism in the connection between feature groups.This gating mechanism dynamically selects channel-wise features based on the input, to suppress the less relevant channels and enhance the detection generalizability.Three gating mechanisms with different structures are proposed and integrated into Res2Net.Experimental results conducted on ASVspoof 2019 logical access (LA) demonstrate that the proposed CG-Res2Net significantly outperforms Res2Net on both the overall LA evaluation set and individual difficult unseen attacks, which also outperforms other state-of-the-art single systems, depicting the effectiveness of our method.

This paper describes a variational auto-encoder based nonautoregressive text-to-speech (VAENAR-TTS) model.The autoregressive TTS (AR-TTS) models based on the sequenceto-sequence architecture can generate high-quality speech, but their sequential decoding process can be time-consuming.Recently, non-autoregressive TTS (NAR-TTS) models have been shown to be more efficient with the parallel decoding process.However, these NAR-TTS models rely on phoneme-level durations to generate a hard alignment between the text and the spectrogram.Obtaining duration labels, either through forced alignment or knowledge distillation, is cumbersome.Furthermore, hard alignment based on phoneme expansion can degrade the naturalness of the synthesized speech.In contrast, the proposed model of VAENAR-TTS is an end-to-end approach that does not require phoneme-level durations.The VAENAR-TTS model does not contain recurrent structures and is completely non-autoregressive in both the training and inference phases.Based on the VAE architecture, the alignment information is encoded in the latent variable, and attention-based soft alignment between the text and the latent variable is used in the decoder to reconstruct the spectrogram.Experiments show that VAENAR-TTS achieves state-of-the-art synthesis quality, while the synthesis speed is comparable with other NAR-TTS models.

One-shot voice conversion (VC), which performs conversion across arbitrary speakers with only a single target-speaker utterance for reference, can be effectively achieved by speech representation disentanglement.Existing work generally ignores the correlation between different speech representations during training, which causes leakage of content information into the speaker representation and thus degrades VC performance.To alleviate this issue, we employ vector quantization (VQ) for content encoding and introduce mutual information (MI) as the correlation metric during training, to achieve proper disentanglement of content, speaker and pitch representations, by reducing their inter-dependencies in an unsupervised manner.Experimental results reflect the superiority of the proposed method in learning effective disentangled speech representations for retaining source linguistic content and intonation variations, while capturing target speaker characteristics.In doing so, the proposed approach achieves higher speech naturalness and speaker similarity than current state-of-the-art one-shot VC systems.Our code, pre-trained models and demo are available at https://github.com/Wendison/VQMIVC.

Dysarthric speech detection (DSD) systems aim to detect characteristics of the neuromotor disorder from speech.Such systems are particularly susceptible to domain mismatch where the training and testing data come from the source and target domains respectively, but the two domains may differ in terms of speech stimuli, disease etiology, etc.It is hard to acquire labelled data in the target domain, due to high costs of annotating sizeable datasets.This paper makes a first attempt to formulate cross-domain DSD as an unsupervised domain adaptation (UDA) problem.We use labelled source-domain data and unlabelled target-domain data, and propose a multi-task learning strategy, including dysarthria presence classification (DPC), domain adversarial training (DAT) and mutual information minimization (MIM), which aim to learn dysarthriadiscriminative and domain-invariant biomarker embeddings.Specifically, DPC helps biomarker embeddings capture critical indicators of dysarthria; DAT forces biomarker embeddings to be indistinguishable in source and target domains; and MIM further reduces the correlation between biomarker embeddings and domain-related cues.By treating the UASPEECH and TORGO corpora respectively as the source and target domains, experiments show that the incorporation of UDA attains absolute increases of 22.2% and 20.0% respectively in utterancelevel weighted average recall and speaker-level accuracy.

Though significant progress has been made for the voice conversion (VC) of typical speech, VC for atypical speech, e.g., dysarthric and second-language (L2) speech, remains a challenge, since it involves correcting for atypical prosody while maintaining speaker identity.To address this issue, we propose a VC system with explicit prosodic modelling and deep speaker embedding (DSE) learning.First, a speech-encoder strives to extract robust phoneme embeddings from atypical speech.Second, a prosody corrector takes in phoneme embeddings to infer typical phoneme duration and pitch values.Third, a conversion model takes phoneme embeddings and typical prosody features as inputs to generate the converted speech, conditioned on the target DSE that is learned via speaker encoder or speaker adaptation.Extensive experiments demonstrate that speaker adaptation can achieve higher speaker similarity, and the speaker encoder based conversion model can greatly reduce dysarthric and non-native pronunciation patterns with improved speech intelligibility.A comparison of speech recognition results between the original dysarthric speech and converted speech show that absolute reduction of 47.6% character error rate (CER) and 29.3% word error rate (WER) can be achieved.

Dysarthric speech recognition is a challenging task due to acoustic variability and limited amount of available data. Diverse conditions of dysarthric speakers account for the acoustic variability, which make the variability difficult to be modeled precisely. This paper presents a variational auto-encoder based variability encoder (VAEVE) to explicitly encode such variability for dysarthric speech. The VAEVE makes use of both phoneme information and low-dimensional latent variable to reconstruct the input acoustic features, thereby the latent variable is forced to encode the phoneme-independent variability. Stochastic gradient variational Bayes algorithm is applied to model the distribution for generating variability encodings, which are further used as auxiliary features for DNN acoustic modeling. Experiment results conducted on the UASpeech corpus show that the VAEVE based variability encodings have complementary effect to the learning hidden unit contributions (LHUC) speaker adaptation. The systems using variability encodings consistently outperform the comparable baseline systems without using them, and" obtain absolute word error rate (WER) reduction by up to 2.2% on dysarthric speech with "Very lowintelligibility level, and up to 2% on the "Mixed" type of dysarthric speech with diverse or uncertain conditions.

Discriminative training techniques define state-of-the-art performance for automatic speech recognition systems. However, they are inherently prone to overfitting, leading to poor generalization performance when using limited training data. In order to address this issue, this paper presents a full Bayesian framework to account for model uncertainty in sequence discriminative training of factored TDNN acoustic models. Several Bayesian learning based TDNN variant systems are proposed to model the uncertainty over weight parameters and choices of hidden activation functions, or the hidden layer outputs. Efficient variational inference approaches using as few as one single parameter sample ensure their computational cost in both training and evaluation time comparable to that of the baseline TDNN systems. Statistically significant word error rate (WER) reductions of 0.4%-1.8% absolute (5%-11% relative) were obtained over a state-of-the-art 900 h speed perturbed Switchboard corpus trained baseline LF-MMI factored TDNN system using multiple regularization methods including F-smoothing, L2 norm penalty, natural gradient, model averaging and dropout, in addition to i-Vector plus learning hidden unit contribution (LHUC) based speaker adaptation and RNNLM rescoring. The efficacy of the proposed Bayesian techniques is further demonstrated in a comparison against the state-of-the-art performance obtained on the same task using the most recent hybrid and end-to-end systems reported in the literature. Consistent performance improvements were also obtained on a 450-h HKUST conversational Mandarin telephone speech recognition task. On a third cross domain adaptation task requiring rapidly porting a 1000-h LibriSpeech data trained system to a small DementiaBank elderly speech corpus, the proposed Bayesian TDNN LF-MMI systems outperformed the baseline system using direct weight fine-tuning by up to 2.5% absolute WER reduction.

This paper proposes an any-to-many location-relative, sequence-to-sequence (seq2seq), non-parallel voice conversion approach, which utilizes text supervision during training. In this approach, we combine a bottle-neck feature extractor (BNE) with a seq2seq synthesis module. During the training stage, an encoder-decoder-based hybrid connectionist-temporal-classification-attention (CTC-attention) phoneme recognizer is trained, whose encoder has a bottle-neck layer. A BNE is obtained from the phoneme recognizer and is utilized to extract speaker-independent, dense and rich spoken content representations from spectral features. Then a multi-speaker location-relative attention based seq2seq synthesis model is trained to reconstruct spectral features from the bottle-neck features, conditioning on speaker representations for speaker identity control in the generated speech. To mitigate the difficulties of using seq2seq models to align long sequences, we down-sample the input spectral feature along the temporal dimension and equip the synthesis model with a discretized mixture of logistic (MoL) attention mechanism. Since the phoneme recognizer is trained with large speech recognition data corpus, the proposed approach can conduct any-to-many voice conversion. Objective and subjective evaluations show that the proposed any-to-many approach has superior voice conversion performance in terms of both naturalness and speaker similarity. Ablation studies are conducted to confirm the effectiveness of feature selection and model design strategies in the proposed approach. The proposed VC approach can readily be extended to support any-to-any VC (also known as one/few-shot VC), and achieve high performance according to objective and subjective evaluations.

Despite the rapid progress of automatic speech recognition (ASR) technologies in the past few decades, recognition of disordered speech remains a highly challenging task to date. Disordered speech presents a wide spectrum of challenges to current data intensive deep neural networks (DNNs) based ASR technologies that predominantly target normal speech. This paper presents recent research efforts at the Chinese University of Hong Kong (CUHK) to improve the performance of disordered speech recognition systems on the largest publicly available UASpeech dysarthric speech corpus. A set of novel modelling techniques including neural architectural search, data augmentation using spectra-temporal perturbation, model based speaker adaptation and cross-domain generation of visual features within an audio-visual speech recognition (AVSR) system framework were employed to address the above challenges. The combination of these techniques produced the lowest published word error rate (WER) of 25.21% on the UASpeech test set 16 dysarthric speakers, and an overall WER reduction of 5.4% absolute (17.6% relative) over the CUHK 2018 dysarthric speech recognition system featuring a 6-way DNN system combination and cross adaptation of out-of-domain normal speech data trained systems. Bayesian model adaptation further allows rapid adaptation to individual dysarthric speakers to be performed using as little as 3.06 seconds of speech. The efficacy of these techniques were further demonstrated on a CUDYS Cantonese dysarthric speech recognition task.

Expressive text-to-speech (E-TTS) synthesis is important for enhancing user experience in communication with machines using the speech modality. However, one of the challenges in E-TTS is the lack of a precise description of emotions. Previous categorical specifications may be insufficient for describing complex emotions. The dimensional specifications face the difficulty of ambiguity in annotation. This work advocates a new approach of describing emotive speech acoustics using spoken exemplars. We investigate methods to extract emotion descriptions from the input exemplar of emotive speech. The measures are combined to form two descriptors, based on capsule network (CapNet) and residual error network (RENet). The first is designed to consider the spatial information in the input exemplary spectrogram, and the latter is to capture the contrastive information between emotive acoustic expressions. Two different approaches are applied for conversion from the variable-length feature sequence to fixed-size description vector: (1) dynamic routing groups similar capsules to the output description; and (2) recurrent neural network's hidden states store the temporal information for the description. The two descriptors are integrated to a state-of-the-art sequence-to-sequence architecture to obtain an end-to-end architecture that is optimized as a whole towards the same goal of generating correct emotive speech. Experimental results on a public audiobook dataset demonstrate that the two exemplar-based approaches achieve significant performance improvement over the baseline system in both emotion similarity and speech quality.

Speech emotion recognition (SER) is an indispensable part of fluid human-machine interaction and attracts lots of research attentions. Recent work on SER has successfully applied convolutional neural networks (CNNs) to learn feature representations from speech spectrograms. However, the fundamental problem of CNNs is that the spatial information in spectrograms is lost, which includes positional and relationship information of low-level features, such as pitch and formant frequencies. We propose a novel architecture of sequential capsule networks (CapNets) by leveraging the advantange of CapNets that spatial information can be preserved in capsules and passed to upper capsule layers via dynamic routing. Also, the dynamic routing algorithm provides an effective alternative to pooling or storing recurrent hidden states for obtaining utterance-level features from the sequential capsule outputs. To further improve the model's ability to capture contextual information, we introduce a recurrent connection to the sequential structure. The experimental comparison of the proposed systems and previously published systems using CNNs and recurrent neural networks (RNNs) based on the IEMOCAP corpus demonstrates the effectiveness of the proposed sequential CapNets.

A key task for speech recognition systems is to reduce the mismatch between training and evaluation data that is often attributable to speaker differences. Speaker adaptation techniques play a vital role to reduce the mismatch. Model-based speaker adaptation approaches often require sufficient amounts of target speaker data to ensure robustness. When the amount of speaker level data is limited, speaker adaptation is prone to overfitting and poor generalization. To address the issue, this paper proposes a full Bayesian learning based DNN speaker adaptation framework to model speaker-dependent (SD) parameter uncertainty given limited speaker specific adaptation data. This framework is investigated in three forms of model based DNN adaptation techniques: Bayesian learning of hidden unit contributions (BLHUC), Bayesian parameterized activation functions (BPAct), and Bayesian hidden unit bias vectors (BHUB). In the three methods, deterministic SD parameters are replaced by latent variable posterior distributions for each speaker, whose parameters are efficiently estimated using a variational inference based approach. Experiments conducted on 300-hour speed perturbed Switchboard corpus trained LF-MMI TDNN/CNN-TDNN systems suggest the proposed Bayesian adaptation approaches consistently outperform the deterministic adaptation on the NIST Hub5'00 and RT03 evaluation sets. When using only the first five utterances from each speaker as adaptation data, significant word error rate reductions up to 1.4% absolute (7.2% relative) were obtained on the CallHome subset. The efficacy of the proposed Bayesian adaptation techniques is further demonstrated in a comparison against the state-of-the-art performance obtained on the same task using the most recent systems reported in the literature.

State-of-the-art language models (LMs) represented by long-short term memory recurrent neural networks (LSTM-RNNs) and Transformers are becoming increasingly complex and expensive for practical applications. Low-bit neural network quantization provides a powerful solution to dramatically reduce their model size. Current quantization methods are based on uniform precision and fail to account for the varying performance sensitivity at different parts of LMs to quantization errors. To this end, novel mixed precision neural network LM quantization methods are proposed in this paper. The optimal local precision choices for LSTM-RNN and Transformer based neural LMs are automatically learned using three techniques. The first two approaches are based on quantization sensitivity metrics in the form of either the KL-divergence measured between full precision and quantized LMs, or Hessian trace weighted quantization perturbation that can be approximated efficiently using matrix free techniques. The third approach is based on mixed precision neural architecture search. In order to overcome the difficulty in using gradient descent methods to directly estimate discrete quantized weights, alternating direction methods of multipliers (ADMM) are used to efficiently train quantized LMs. Experiments were conducted on state-of-the-art LF-MMI CNN-TDNN systems featuring speed perturbation, i-Vector and learning hidden unit contribution (LHUC) based speaker adaptation on two tasks: Switchboard telephone speech and AMI meeting transcription. The proposed mixed precision quantization techniques achieved “lossless” quantization on both tasks, by producing model size compression ratios of up to approximately 16 times over the full precision LSTM and Transformer baseline LMs, while incurring no statistically significant word error rate increase.

Automatic speech recognition (ASR) technologies have been significantly advanced in the past few decades. However, recognition of overlapped speech remains a highly challenging task to date. To this end, multi-channel microphone array data are widely used in current ASR systems. Motivated by the invariance of visual modality to acoustic signal corruption and the additional cues they provide to separate the target speaker from the interfering sound sources, this paper presents an audio-visual multi-channel based recognition system for overlapped speech. It benefits from a tight integration between a speech separation front-end and recognition back-end, both of which incorporate additional video input. A series of audio-visual multi-channel speech separation front-end components based on TF masking, Filter&Sum and mask-based MVDR neural channel integration approaches are developed. To reduce the error cost mismatch between the separation and the recognition components, the entire system is jointly fine-tuned using a multi-task criterion interpolation of the scale-invariant signal to noise ratio (Si-SNR) with either the connectionist temporal classification (CTC), or lattice-free maximum mutual information (LF-MMI) loss function. Experiments suggest that: the proposed audio-visual multi-channel recognition system outperforms the baseline audio-only multi-channel ASR system by up to 8.04% (31.68% relative) and 22.86% (58.51% relative) absolute WER reduction on overlapped speech constructed using either simulation or replaying of the LRS2 dataset respectively. Consistent performance improvements are also obtained using the proposed audio-visual multi-channel recognition system when using occluded video input with the lip region randomly covered up to 60%.

If NAS methods are solutions, what is the problem? Most existing NAS methods require two-stage parameter optimization. However, performance of the same architecture in the two stages correlates poorly. In this work, we propose a new problem definition for NAS, task-specific end-to-end, based on this observation. We argue that given a computer vision task for which a NAS method is expected, this definition can reduce the vaguely-defined NAS evaluation to i) accuracy of this task and ii) the total computation consumed to finally obtain a model with satisfying accuracy. Seeing that most existing methods do not solve this problem directly, we propose DSNAS, an efficient differentiable NAS framework that simultaneously optimizes architecture and parameters with a low-biased Monte Carlo estimate. Child networks derived from DSNAS can be deployed directly without parameter retraining. Comparing with two-stage methods, DSNAS successfully discovers networks with comparable accuracy (74.4\%) on ImageNet in 420 GPU hours, reducing the total time by more than 34\%.

Synthesizing fluent code-switched (CS) speech with consistent voice using only monolingual corpora is still a challenging task, since language alternation seldom occurs during training and the speaker identity is directly correlated with language. In this paper, we present a bilingual phonetic posteriorgram (PPG) based CS speech synthesizer using only monolingual corpora. The bilingual PPG is used to bridge across speakers and languages, which is formed by stacking two monolingual PPGs extracted from two monolingual speaker-independent speech recognition systems. It is assumed that bilingual PPG can represent the articulation of speech sounds speaker-independently and captures accurate phonetic information of both languages in the same feature space. The proposed model first extracts bilingual PPGs from training data. Then an encoder- decoder based model is used to learn the relationship between input text and bilingual PPGs, and the bilingual PPGs are mapped to acoustic features using bidirectional long-short term memory based model conditioned on speaker embedding to control speaker identity. Experiments validate the effectiveness of the proposed model in terms of speech intelligibility, audio fidelity and speaker consistency of the generated code-switched speech.

This work investigates the vulnerability of Gaussian Mixture Model (GMM) i-vector based speaker verification systems to adversarial attacks, and the transferability of adversarial samples crafted from GMM i-vector based systems to x-vector based systems. In detail, we formulate the GMM i-vector system as a scoring function of enrollment and testing utterance pairs. Then we leverage the fast gradient sign method (FGSM) to optimize testing utterances for adversarial samples generation. These adversarial samples are used to attack both GMM i-vector and x-vector systems. We measure the system vulnerability by the degradation of equal error rate and false acceptance rate. Experiment results show that GMM i-vector systems are seriously vulnerable to adversarial attacks, and the crafted adversarial samples are proved to be transferable and pose threats to neural network speaker embedding based systems (e.g. x-vector systems).

Techniques for accent conversion (AC) aim to convert non-native to native accented speech. Conventional AC methods try to convert only the speaker identity of a native speaker's voice to that of the non-native accented target speaker, leaving the underlying content and pronunciations unchanged. This hinders their practical use in real-world applications, because native-accented utterances are required at conversion stage. In this paper, we present an end-to-end framework, which is able to conduct AC from non-native-accented utterances without using any native-accented utterances during online conversion. We achieve this by independently extracting linguistic and speaker representations from non-native accented speech and condition a speech synthesis model on these representations to generate native-accented speech. Experiments on open-source data corpora show that the proposed system can convert Hindi-accented English speech into native American English speech with high naturalness, which is indistinguishable from native-accented recordings in terms of accent.

Dysarthric speech reconstruction (DSR) is a challenging task due to difficulties in repairing unstable prosody and correcting imprecise articulation. Inspired by the success of sequence-to-sequence (seq2seq) based text-to-speech (TTS) synthesis and knowledge distillation (KD) techniques, this paper proposes a novel end-to-end voice conversion (VC) method to tackle the reconstruction task. The proposed approach contains three components. First, a seq2seq based TTS is first trained with the transcribed normal speech. Second, with the text-encoder of this trained TTS system as "teacher", a teacher-student framework is proposed for cross-modal KD by training a speech-encoder to extract appropriate linguistic representations from the transcribed dysarthric speech. Third, the speech-encoder of the previous component is concatenated with the attention and decoder of the first component (TTS) to perform the DSR task, by directly mapping the dysarthric speech to its normal version. Experiments demonstrate that the proposed method can generate the speech with high naturalness and intelligibility, where the comparisons of human speech recognition between the reconstructed speech and the original dysarthric speech show that 35.4% and 48.7% absolute word error rate (WER) reduction can be achieved for dysarthric speakers with low and very low speech intelligibility, respectively.

The high memory consumption and computational costs of Recurrent neural network language models (RNNLMs) limit their wider application on resource constrained devices. In recent years, neural network quantization techniques that are capable of producing extremely low-bit compression, for example, binarized RNNLMs, are gaining increasing research interests. Directly training of quantized neural networks is difficult. By formulating quantized RNNLMs training as an optimization problem, this paper presents a novel method to train quantized RNNLMs from scratch using alternating direction methods of multipliers (ADMM). This method can also flexibly adjust the trade-off between the compression rate and model performance using tied low-bit quantization tables. Experiments on two tasks: Penn Treebank (PTB), and Switchboard (SWBD) suggest the proposed ADMM quantization achieved a model size compression factor of up to 31 times over the full precision baseline RNNLMs. Faster convergence of 5 times in model training over the baseline binarized RNNLM quantization was also obtained.

Automatic recognition of overlapped speech remains a highly challenging task to date. Motivated by the bimodal nature of human speech perception, this paper investigates the use of audio-visual technologies for overlapped speech recognition. Three issues associated with the construction of audio-visual speech recognition (AVSR) systems are addressed. First, the basic architecture designs i.e. end-to-end and hybrid of AVSR systems are investigated. Second, purposefully designed modality fusion gates are used to robustly integrate the audio and visual features. Third, in contrast to a traditional pipelined architecture containing explicit speech separation and recognition components, a streamlined and integrated AVSR system optimized consistently using the lattice-free MMI (LF-MMI) discriminative criterion is also proposed. The proposed LF-MMI time-delay neural network (TDNN) system establishes the state-of-the-art for the LRS2 dataset. Experiments on overlapped speech simulated from the LRS2 dataset suggest the proposed AVSR system outperformed the audio only baseline LF-MMI DNN system by up to 29.98% absolute in word error rate (WER) reduction, and produced recognition performance comparable to a more complex pipelined system. Consistent performance improvements of 4.89% absolute in WER reduction over the baseline AVSR system using feature fusion are also obtained.

Disordered speech recognition is a highly challenging task. The underlying neuro-motor conditions of people with speech disorders, often compounded with co-occurring physical disabilities, lead to the difficulty in collecting large quantities of speech required for system development. This paper investigates a set of data augmentation techniques for disordered speech recognition, including vocal tract length perturbation (VTLP), tempo perturbation and speed perturbation. Both normal and disordered speech were exploited in the augmentation process. Variability among impaired speakers in both the original and augmented data was modeled using learning hidden unit contributions (LHUC) based speaker adaptive training. The final speaker adapted system constructed using the UASpeech corpus and the best augmentation approach based on speed perturbation produced up to 2.92% absolute (9.3% relative) word error rate (WER) reduction over the baseline system without data augmentation, and gave an overall WER of 26.37% on the test set containing 16 dysarthric speakers.

Recently adversarial attacks on automatic speaker verification (ASV) systems attracted widespread attention as they pose severe threats to ASV systems.However, methods to defend against such attacks are limited.Existing approaches mainly focus on retraining ASV systems with adversarial data augmentation.Also, countermeasure robustness against different attack settings are insufficiently investigated.Orthogonal to prior approaches, this work proposes to defend ASV systems against adversarial attacks with a separate detection network, rather than augmenting adversarial data into ASV training.A VGG-like binary classification detector is introduced and demonstrated to be effective on detecting adversarial samples.To investigate detector robustness in a realistic defense scenario where unseen attack settings may exist, we analyze various kinds of unseen attack settings' impact and observe that the detector is robust (6.27%EER det degradation in the worst case) against unseen substitute ASV systems, but it has weak robustness (50.37%EER det degradation in the worst case) against unseen perturbation methods.The weak robustness against unseen perturbation methods shows a direction for developing stronger countermeasures.

Voice conversion (VC) techniques aim to modify speaker identity of an utterance while preserving the underlying linguistic information.Most VC approaches ignore modeling of the speaking style (e.g.emotion and emphasis), which may contain the factors intentionally added by the speaker and should be retained during conversion.This study proposes a sequence-tosequence based non-parallel VC approach, which has the capability of transferring the speaking style from the source speech to the converted speech by explicitly modeling.Objective evaluation and subjective listening tests show superiority of the proposed VC approach in terms of speech naturalness and speaker similarity of the converted speech.Experiments are also conducted to show the source-style transferability of the proposed approach.

Automatic speech recognition (ASR) of overlapped speech remains a highly challenging task to date. To this end, multi-channel microphone array data are widely used in state-of-the-art ASR systems. Motivated by the invariance of visual modality to acoustic signal corruption, this paper presents an audio-visual multi-channel overlapped speech recognition system featuring tightly integrated separation front-end and recognition back-end. A series of audio-visual multi-channel speech separation front-end components based on \textit{TF masking}, \textit{filter\&sum} and \textit{mask-based MVDR} beamforming approaches were developed. To reduce the error cost mismatch between the separation and recognition components, they were jointly fine-tuned using the connectionist temporal classification (CTC) loss function, or a multi-task criterion interpolation with scale-invariant signal to noise ratio (Si-SNR) error cost. Experiments suggest that the proposed multi-channel AVSR system outperforms the baseline audio-only ASR system by up to 6.81\% (26.83\% relative) and 22.22\% (56.87\% relative) absolute word error rate (WER) reduction on overlapped speech constructed using either simulation or replaying of the lipreading sentence 2 (LRS2) dataset respectively.

There has been a long term interest in using visual information to improve automatic speech recognition (ASR) system performance. Both audio and visual information are required in conventional audio visual speech recognition (AVSR) systems. This limits their wider applications when visual modality is not present. To this end, one possible solution is to use acoustic-to-visual (A2V) inversion techniques to generate visual features. Previous research in this direction used synthetic acoustic-articulatory parallel data in inversion model training. The acoustic mismatch between the audio-visual (AV) parallel data and target data was not considered. In addition, the target language to apply these technologies has been focused on English. In this article, a real 3D Audio-Visual Mandarin Continuous Speech (3DAV-MCS) corpus was used to train deep neural network based A2V inversion models. Cross-domain adaptation of the inversion models allows suitable visual features to be generated from acoustic data of mismatched domains. The proposed cross-domain deep visual feature generation techniques were evaluated on two state-of-the-art Mandarin speech recognition tasks: DAPRA GALE broadcast transcription and BOLT conversational telephone speech recognition. The AVSR systems constructed using the cross-domain generated visual features consistently outperformed the baseline convolutional neural network (CNN) ASR systems by up to 3.3% absolute (9.1% relative) character error rate (CER) reductions after both speaker adaptive training and sequence discriminative training were performed.

State-of-the-art text-to-speech (TTS) synthesis models can produce monolingual speech with high intelligibility and naturalness. However, when the models are applied to synthesize code-switched (CS) speech, the performance declines seriously. Conventionally, developing a CS TTS system requires multilingual data to incorporate language-specific and cross-lingual knowledge. Recently, end-to-end (E2E) architecture has achieved satisfactory results in monolingual TTS. The architecture enables the training from one end of alphabetic text input to the other end of acoustic feature output. In this paper, we explore the use of E2E framework for CS TTS, using a combination of Mandarin and English monolingual speech corpus uttered by two female speakers. To handle alphabetic input from different languages, we explore two kinds of encoders: (1) shared multilingual encoder with explicit language embedding (LDE); (2) separated monolingual encoder (SPE) for each language. The two systems use identical decoder architecture, where a discriminative code is incorporated to enable the model to generate speech in one speaker's voice consistently. Experiments confirm the effectiveness of the proposed modifications on the E2E TTS framework in terms of quality and speaker similarity of the generated speech. Moreover, our proposed systems can generate controllable foreign-accented speech at character-level using only mixture of monolingual training data.

The hidden activation functions inside deep neural networks (DNNs) play a vital role in learning high level discriminative features and controlling the information flows to track longer history. However, the fixed model parameters used in standard DNNs can lead to over-fitting and poor generalization when given limited training data. Furthermore, the precise forms of activations used in DNNs are often manually set at a global level for all hidden nodes, thus lacking an automatic selection method. In order to address these issues, Bayesian neural networks (BNNs) acoustic models are proposed in this paper to explicitly model the uncertainty associated with DNN parameters. Gaussian Process (GP) activations based DNN and LSTM acoustic models are also used in this paper to allow the optimal forms of hidden activations to be stochastically learned for individual hidden nodes. An efficient variational inference based training algorithm is derived for BNN, GPNN and GPLSTM systems. Experiments were conducted on a LVCSR system trained on a 75 hour subset of Switchboard I data. The best BNN and GPNN systems outperformed both the baseline DNN systems constructed using fixed form activations and their combination via frame level joint decoding by 1% absolute in word error rate.

Recurrent neural network language models (RNNLMs) have shown superior performance across a range of speech recognition tasks. At the heart of all RNNLMs, the activation functions play a vital role to control the information flows and tracking longer history contexts that are useful for predicting the following words. Long short-term memory (LSTM) units are well known for such ability and thus widely used in current RNNLMs. However, the deterministic parameter estimates in LSTM RNNLMs are prone to over-fitting and poor generalization when given limited training data. Furthermore, the precise forms of activations in LSTM have been largely empirically set for all cells at a global level. In order to address these issues, this paper introduces Gaussian process (GP) LSTM RNNLMs. In addition to modeling parameter uncertainty under a Bayesian framework, it also allows the optimal forms of gates being automatically learned for individual LSTM cells. Experiments were conducted on three tasks: the Penn Treebank (PTB) corpus, Switchboard conversational telephone speech (SWBD) and the AMI meeting room data. The proposed GP-LSTM RNNLMs consistently outperform the baseline LSTM RNNLMs in terms of both perplexity and word error rate.

This paper focuses on using Convolutional Neural Network (CNN), Recurrent Neural Network (RNN) and Connection-ist Temporal Classification (CTC) to build an end-to-end speech recognition for Mispronunciation Detection and Diagnosis (MDD) task. Our approach is end-to-end models, while phonemic or graphemic information, or forced alignment between different linguistic units, are not required. We conduct experiments that compare the proposed CNN-RNN-CTC approach with alternative mispronunciation detection and diagnoses (MDD) approaches. The F-measure of our approach is 74.65%, which significantly outperforms the Extended Recognition Network (ERN) (S-AM) by 44.75% and State-level Acoustic Model (S-AM) by 32.28% relatively. The relative improvement in F-measure when over Acoustic-Phonemic Model (APM), Acoustic-Graphemic Model (AGM) and Acoustic-Phonemic-Graphemic Model (APGM) are 9.57%, 5.04% and 2.77% respectively.

Speech emotion recognition (SER) is a fundamental step towards fluent human-machine interaction. One challenging problem in SER is obtaining utterance-level feature representation for classification. Recent works on SER have made significant progress by using spectrogram features and introducing neural network methods, e.g., convolutional neural networks (CNNs). However the fundamental problem of CNNs is that the spatial information in spectrograms is not captured, which are basically position and relationship information of low-level features like pitch and formant frequencies. This paper presents a novel architecture based on the capsule networks (CapsNets) for SER. The proposed system can take into account the spatial relationship of speech features in spectrograms, and provide an effective pooling method for obtaining utterance global features. We also introduce a recurrent connection to CapsNets to improve the model's time sensitivity. We compare the proposed model to previous published results based on combined CNN-long short-term memory (CNN-LSTM) models on the benchmark corpus IEMOCAP over four emotions, i.e., neutral, angry, happy and sad. Experimental results show that our model achieves better results than the baseline system on weighted accuracy (WA) (72.73% vs. 68.8%) and un-weighted accuracy (UA) (59.71% vs. 59.4%), which demonstrates the effectiveness of CapsNets for SER.

Speaker adaptation techniques play a key role in reducing the mismatch between speech recognition systems and target users. In order to robustly learn speaker-dependent adaptation parameters, model based DNN adaptation techniques often require a significant amount of data. For example, in the commonly used learning hidden unit contributions (LHUC) based DNN adaptation, speaker-dependent high-dimensional hidden layer output scaling vectors are used. When limited adaptation data are available, the standard L-HUC is prone to over-fitting and poor generalization. To address the issue, Bayesian learning of hidden unit contributions (BLHUC) is proposed in this paper. A posterior distribution over the LHUC scaling vectors is used to explicitly model the uncertainty associated with the adaptation parameters. An efficient variational inference based approach is adopted to estimate the LHUC parameter posterior distribution. Experiments conducted on a 300-hour Switchboard setup showed that the proposed BLHUC method outperformed the baseline speaker-independent DNN systems and LHUC adapted DNN systems by up to 1.4% and 1.1% absolute reductions of word error rate respectively, when only using 1 utterance of adaptation data from each speaker. Consistent performance improvements were also obtained over the baseline, LHUC adapted and LHUC SAT systems when increasing the amount of adaptation data.

Recurrent neural network language models (RNNLMs) have become an increasing popular choice for state-of-the-art speech recognition systems. RNNLMs are normally trained by minimizing the cross entropy (CE) using the stochastic gradient descent (SGD) algorithm. However, the SGD method doesn't consider the correlation between parameters and therefore can lead to unstable and slow convergence in training. Second-order optimization methods provide a possible solution to this issue. However these methods are either computationally heavy or do not have competitive performance. In this paper, a novel optimization method - stochastic natural gradient based on minimum variance assumption (SNGM) is proposed for training RNNLMs. It allows the natural gradient method to operate at a comparable training efficiency to the SGD method. By modifying the gradient according to the local curvature of the KL-divergence between current and updated probabilistic distributions, the proposed SNGM approach is shown to outperform both the SGD and limited memory BFGS methods across three tasks: Penn Treebank, Switchboard conversational speech recognition and AMI meeting room transcription in terms of both perplexity and word error rate.

Language modeling is a crucial component in a wide range of applications including speech recognition. Language models (LMs) are usually constructed by splitting a sentence into words and computing the probability of a word based on its word history. This sentence probability calculation, making use of conditional probability distributions, assumes that there is little impact from approximations used in the LMs, including the word history representations and finite training data. This motivates examining models that make use of additional information from the sentence. In this paper, future word information, in addition to the history, is used to predict the probability of the current word. For recurrent neural network LMs (RNNLMs), this information can be encapsulated in a bi-directional model. However, if used directly, this form of model is computationally expensive when trained on large quantities of data, and can be problematic when used with word lattices. This paper proposes a novel neural network language model structure, the succeeding-word RNNLM, su-RNNLM, to address these issues. Instead of using a recurrent unit to capture the complete future word contexts, a feedforward unit is used to model a fixed finite number of succeeding words. This is more efficient in training than bi-directional models and can be applied to lattice rescoring. The generated lattices can be used for downstream applications, such as confusion network decoding and keyword search. Experimental results on speech recognition and keyword spotting tasks illustrate the empirical usefulness of future word information, and the flexibility of the proposed model to represent this information.

Recurrent neural network language models (RNNLM) have become an increasingly popular choice for state-of-the-art speech recognition systems. RNNLMs are normally trained by minimizing the cross entropy (CE) using the stochastic gradient descent (SGD) algorithm. The SGD method only uses first-order derivatives and no higher order gradient information is used to consider the correlation between model parameters. It is unable to fully capture the curvature of the error cost function. This can lead to slow convergence in model training. In this paper, a limited-memory Broyden Fletcher Goldfarb Shannon (L-BFGS) based second order optimization technique is proposed for RNNLMs. This method efficiently approximates the matrix-vector product between the inverse Hessian and gradient vector via a recursion over past gradients with a compact memory requirement. Consistent perplexity and error rate reductions are obtained over the SGD method on two speech recognition tasks: Switchboard English and Babel Cantonese. A faster convergence and speed up in RNNLM training time was also obtained.

Second language (L2) speech is often labelled with the native, phoneme categories. Hence, we often observe segments for which it is difficult, if not impossible, to decide on a categorical phoneme label. We refer to these segments as “non-categorical” phoneme units. Existing approaches to mispronunciation detection and diagnosis (MDD) mostly focus on categorical phoneme errors, where one native phoneme is substituted for another. However, noncategorical errors are not considered. To better represent L2 speech for improved MDD, this work aims to discover an Extended Phoneme Set in L2 speech (L2-EPS) which includes not only the categorical phonemes based on the native set, but also non-categorical phoneme units. We apply an optimized k-means algorithm to cluster phoneme-based phonemic posterior-grams (PPGs), which are generated through an acoustic-phonemic model (APM). Then we find the L2-EPS based on analysis of the clusters obtained. We verified experimentally that the non-categorical phonemes in L2-EPS can extend the native phoneme categories to better describe L2 speech. Hence L2-EPS can enrich the existing approaches to MDD for better performance.

Speaking style plays an important role in the expressivity of speech for communication. Hence speaking style is very important for synthetic speech as well. Speaking style adaptation faces the difficulty that the data of specific styles may be limited and difficult to obtain in large amounts. A possible solution is to leverage data from speaking styles that are more available, to train the speech synthesizer and then adapt it to the target style for which the data is scarce. Conventional DNN adaptation approaches directly update the top layers of a well-trained, style-dependent model towards the target style. The detailed local context-level mismatch between the original and the target styles is not considered. In order to address this issue, two frame-level input feature-based style adaptation techniques are investigated in this paper. We will use style features extracted from (1) a target-style data trained bottleneck DNN, and (2) a novel cross-style residual feature regression DNN. These features are used for top-layer adaptation of a well-trained style-dependent synthesis network. Experimental results on adapting the declarative sty le to the interrogative sty le demonstrate the effectiveness of our proposed style features in improving the expressiveness of synthesizing speech for the interrogative style, while maintaining speech quality.

Recurrent neural network language models (RNNLMs) have becoming increasingly popular in many applications such as automatic speech recognition (ASR). Significant performance improvements in both perplexity and word error rate over standard n-gram LMs have been widely reported on ASR tasks. In contrast, published research on using RNNLMs for keyword search systems has been relatively limited. In this paper the application of RNNLMs for the IARPA Babel keyword search task is investigated. In order to supplement the limited acoustic transcription data, large amounts of web texts are also used in large vocabulary design and LM training. Various training criteria were then explored to improved RNNLMs' efficiency in both training and evaluation. Significant and consistent improvements on both keyword search and ASR tasks were obtained across all languages.

Recurrent neural networks (RNNs) and their bidirectional long short term memory (BLSTM) variants are powerful sequence modelling approaches. Their inherently strong ability in capturing long range temporal dependencies allow BLSTM-RNN speech synthesis systems to produce higher quality and smoother speech trajectories than conventional deep neural networks (DNNs). In this paper, we improve the conventional BLSTM-RNN based approach by introducing a multi-task learned structured output layer where spectral parameter targets are conditioned upon pitch parameters prediction. Both objective and subjective experimental results demonstrated the effectiveness of the proposed technique.

Recurrent neural network language models (RNNLMs) are powerful language modeling techniques. Significant performance improvements have been reported in a range of tasks including speech recognition compared to n-gram language models. Conventional n-gram and neural network language models are trained to predict the probability of the next word given its preceding context history. In contrast, bidirectional recurrent neural network based language models consider the context from future words as well. This complicates the inference process, but has theoretical benefits for tasks such as speech recognition as additional context information can be used. However to date, very limited or no gains in speech recognition performance have been reported with this form of model. This paper examines the issues of training bidirectional recurrent neural network language models (bi-RNNLMs) for speech recognition. A bi-RNNLM probability smoothing technique is proposed, that addresses the very sharp posteriors that are often observed in these models. The performance of the bi-RNNLMs is evaluated on three speech recognition tasks: broadcast news; meeting transcription (AMI); and low-resource systems (Babel data). On all tasks gains are observed by applying the smoothing technique to the bi-RNNLM. In addition consistent performance gains can be obtained by combining bi-RNNLMs with n-gram and uni-directional RNNLMs.

There is widespread interest in the relationship between the neurobiological systems supporting human cognition and emerging computational systems capable of emulating these capacities. Human speech comprehension, poorly understood as a neurobiological process, is an important case in point. Automatic Speech Recognition (ASR) systems with near-human levels of performance are now available, which provide a computationally explicit solution for the recognition of words in continuous speech. This research aims to bridge the gap between speech recognition processes in humans and machines, using novel multivariate techniques to compare incremental 'machine states', generated as the ASR analysis progresses over time, to the incremental 'brain states', measured using combined electro- and magneto-encephalography (EMEG), generated as the same inputs are heard by human listeners. This direct comparison of dynamic human and machine internal states, as they respond to the same incrementally delivered sensory input, revealed a significant correspondence between neural response patterns in human superior temporal cortex and the structural properties of ASR-derived phonetic models. Spatially coherent patches in human temporal cortex responded selectively to individual phonetic features defined on the basis of machine-extracted regularities in the speech to lexicon mapping process. These results demonstrate the feasibility of relating human and ASR solutions to the problem of speech recognition, and suggest the potential for further studies relating complex neural computations in human speech comprehension to the rapidly evolving ASR systems that address the same problem domain.

In recent years, recurrent neural network language models (RNNLMs) have become increasingly popular for a range of applications including speech recognition. However, the training of RNNLMs is computationally expensive, which limits the quantity of data, and size of network, that can be used. In order to fully exploit the power of RNNLMs, efficient training implementations are required. This paper introduces an open-source toolkit, the CUED-RNNLM toolkit, which supports efficient GPU-based training of RNNLMs. RNNLM training with a large number of word level output targets is supported, in contrast to existing tools which used class-based output-targets. Support fotN-best and lattice-based rescoring of both HTK and Kaldi format lattices is included. An example of building and evaluating RNNLMs with this toolkit is presented for a Kaldi based speech recognition system using the AMI corpus. All necessary resources including the source code, documentation and recipe are available online1.

Automatic segmentation is a crucial initial processing step for processing multi-genre broadcast (MGB) audio. It is very challenging since the data exhibits a wide range of both speech types and background conditions with many types of non-speech audio. This paper describes a segmentation system for multi-genre broadcast audio with deep neural network (DNN) based speech/non-speech detection. A further stage of change-point detection and clustering is used to obtain homogeneous segments. Suitable DNN inputs, context window sizes and architectures are studied with a series of experiments using a large corpus of MGB television audio. For MGB transcription, the improved segmenter yields roughly half the increase in word error rate, over manual segmentation, compared to the baseline DNN segmenter supplied for the 2015 ASRU MGB challenge.

This paper compares schemes for the selection of multi-genre broadcast data and corresponding transcriptions for speech recognition model training. Selections of the same amount of data (700 hours) from lightly supervised alignments based on the same original subtitle transcripts are compared. Data segments were selected according to a maximum phone matched error rate between the lightly supervised decoding and the original transcript. The data selected with an improved lightly supervised system yields lower word error rates (WERs). Detailed comparisons of the data selected on carefully transcribed development data show how the selected portions match the true phone error rate for each genre. From a broader perspective, it is shown that for different genres, either the original subtitles or the lightly supervised output should be used for model training and a suitable combination yields further reductions in final WER.

Recurrent neural network language models (RNNLMs) are becoming increasingly popular for a range of applications including automatic speech recognition. An important issue that limits their possible application areas is the computational cost incurred in training and evaluation. This paper describes a series of new efficiency improving approaches that allows RNNLMs to be more efficiently trained on graphics processing units (GPUs) and evaluated on CPUs. First, a modified RNNLM architecture with a nonclass-based, full output layer structure (F-RNNLM) is proposed. This modified architecture facilitates a novel spliced sentence bunch mode parallelization of F-RNNLM training using large quantities of data on a GPU. Second, two efficient RNNLM training criteria based on variance regularization and noise contrastive estimation are explored to specifically reduce the computation associated with the RNNLM output layer softmax normalisation term. Finally, a pipelined training algorithm utilizing multiple GPUs is also used to further improve the training speed. Initially, RNNLMs were trained on a moderate dataset with 20M words from a large vocabulary conversational telephone speech recognition task. The training time of RNNLM is reduced by up to a factor of 53 on a single GPU over the standard CPU-based RNNLM toolkit. A 56 times speed up in test time evaluation on a CPU was obtained over the baseline F-RNNLMs. Consistent improvements in both recognition accuracy and perplexity were also obtained over C-RNNLMs. Experiments on Google's one billion corpus also reveals that the training of RNNLM scales well.

An important part of the language modelling problem for automatic speech recognition (ASR) systems, and many other related applications, is to appropriately model long-distance context dependencies in natural languages. Hence, statistical language models (LMs) that can model longer span history contexts, for example, recurrent neural network language models (RNNLMs), have become increasingly popular for state-of-the-art ASR systems. As RNNLMs use a vector representation of complete history contexts, they are normally used to rescore N-best lists. Motivated by their intrinsic characteristics, two efficient lattice rescoring methods for RNNLMs are proposed in this paper. The first method uses an n-gram style clustering of history contexts. The second approach directly exploits the distance measure between recurrent hidden history vectors. Both methods produced 1-best performance comparable to a 10 k-best rescoring baseline RNNLM system on two large vocabulary conversational telephone speech recognition tasks for US English and Mandarin Chinese. Consistent lattice size compression and recognition performance improvements after confusion network (CN) decoding were also obtained over the prefix tree structured N-best rescoring approach.

This paper describes the Multi-Genre Broadcast (MGB) Challenge at ASRU 2015, an evaluation focused on speech recognition, speaker diarization, and "lightly supervised" alignment of BBC TV recordings. The challenge training data covered the whole range of seven weeks BBC TV output across four channels, resulting in about 1,600 hours of broadcast audio. In addition several hundred million words of BBC subtitle text was provided for language modelling. A novel aspect of the evaluation was the exploration of speech recognition and speaker diarization in a longitudinal setting — i.e. recognition of several episodes of the same show, and speaker diarization across these episodes, linking speakers. The longitudinal tasks also offered the opportunity for systems to make use of supplied metadata including show title, genre tag, and date/time of transmission. This paper describes the task data and evaluation process used in the MGB challenge, and summarises the results obtained.

Recurrent neural network language models (RNNLMs) have become an increasingly popular choice for speech and language processing tasks including automatic speech recognition (ASR). As the generalization patterns of RNNLMs and n-gram LMs are inherently different, RNNLMs are usually combined with n-gram LMs via a fixed weighting based linear interpolation in state-of-the-art ASR systems. However, previous work doesn't fully exploit the difference of modelling power of the RNNLMs and n-gram LMs as n-gram level changes. In order to fully exploit the detailed n-gram level complementary attributes between the two LMs, a back-off based compact representation of n-gram dependent interpolation weights is proposed in this paper. This approach allows weight parameters to be robustly estimated on limited data. Experimental results are reported on the three tasks with varying amounts of training data. Small and consistent improvements in both perplexity and WER were obtained using the proposed interpolation approach over the baseline fixed weighting based linear interpolation.

This paper presents a multi-stage speaker diarisation system with longitudinal Linking developed on BBC multi-genre data for the 2015 Multi-Genre Broadcast (MGB) challenge. The basic speaker diarisation system draws on techniques from the Cambridge March 2005 system with a new deep neural network (DNN)-based speech/non speech segmenter. A newly developed linking stage is next added to the basic diarisation output aiming at the identification of speakers across multiple episodes of the same series. The longitudinal constraint imposes an incremental processing of the episodes, where speaker labels for each episode can be obtained using only material from the episode in question, and those broadcast earlier in time. The nature of the data as well as the longitudinal linking constraint position this diarisation task as a new open-research topic, and a particularly challenging one. Different linking clustering metrics are compared and the lowest within-episode and cross-episode DER scores are achieved on the MGB challenge evaluation set.

We describe the alignment systems developed both for the preparation of data for the Multi-Genre Broadcast (MGB) challenge and for our participation in the transcription and alignment tasks. Captions of varying quality are aligned with the audio of TV shows that range from few minutes long to more than six hours. Lightly supervised decoding is performed on the audio and the output text is aligned with the original text transcript. Reliable split points are found and the resulting text chunks are force-aligned with the corresponding audio segments. Confidence scores are associated with the aligned data. Multiple refinements — including audio segmentation based on deep neural networks (DNNs) and the use of DNN-based acoustic models — were used to improve the performance. The final MGB alignment system had the highest F-measure value on the evaluation data.

We describe the development of our speech-to-text transcription systems for the 2015 Multi-Genre Broadcast (MGB) challenge. Key features of the systems are: a segmentation system based on deep neural networks (DNNs); the use of HTK 3.5 for building DNN-based hybrid and tandem acoustic models and the use of these models in a joint decoding framework; techniques for adaptation of DNN based acoustic models including parameterised activation function adaptation; alternative acoustic models built using Kaldi; and recurrent neural network language models (RNNLMs) and RNNLM adaptation. The same language models were used with both HTK and Kaldi acoustic models and various combined systems built. The final systems had the lowest error rates on the evaluation data.

Recurrent neural network language models (RNNLMs) are becoming increasingly popular for speech recognition. Previously, we have shown that RNNLMs with a full (non-classed) output layer (F-RNNLMs) can be trained efficiently using a GPU giving a large reduction in training time over conventional class-based models (C-RNNLMs) on a standard CPU. However, since test-time RNNLM evaluation is often performed entirely on a CPU, standard F-RNNLMs are inefficient since the entire output layer needs to be calculated for normalisation. In this paper, it is demonstrated that C-RNNLMs can be efficiently trained on a GPU, using our spliced sentence bunch technique which allows good CPU test-time performance (42× speedup over F-RNNLM). Furthermore, the performance of different classing approaches is investigated. We also examine the use of variance regularisation of the softmax denominator for F-RNNLMs and show that it allows F-RNNLMs to be efficiently used in test (56× speedup on a CPU). Finally the use of two GPUs for F-RNNLM training using pipelining is described and shown to give a reduction in training time over a single GPU by a factor of 1.6×.

In recent years recurrent neural network language models (RNNLMs) have been successfully applied to a range of tasks including speech recognition. However, an important issue that limits the quantity of data used, and their possible application areas, is the computational cost in training. A signi??cant part of this cost is associated with the softmax function at the output layer, as this requires a normalization term to be explicitly calculated. This impacts both the training and testing speed, especially when a large output vocabulary is used. To address this problem, noise contrastive estimation (NCE) is explored in RNNLM training. NCE does not require the above normalization during both training and testing. It is insensitive to the output layer size. On a large vocabulary conversational telephone speech recognition task, a doubling in training speed on a GPU and a 56 times speed up in test time evaluation on a CPU were obtained.

Recurrent neural network language models (RNNLM) have become an increasingly popular choice for state-of-the-art speech recognition systems. Linguistic factors in??uencing the realization of surface word sequences, for example, expressive richness, are only implicitly learned by RNNLMs. Observed sentences and their associated alternative paraphrases representing the same meaning are not explicitly related during training. In order to improve context coverage and generalization, paraphrastic RNNLMs are investigated in this paper. Multiple paraphrase variants were automatically generated and used in paraphrastic RNNLM training. Using a paraphrastic multi-level RNNLM modelling both word and phrase sequences, signi??cant error rate reductions of 0.6% absolute and perplexity reduction of 10% relative were obtained over the baseline RNNLM on a large vocabulary conversational telephone speech recognition system trained on 2000 hours of audio and 545 million words of texts. The overall improvement over the baseline n-gram LM was increased from 8.4% to 11.6% relative.

Recurrent neural network language models (RNNLMs) have recently become increasingly popular for many applications including speech recognition. In previous research RNNLMs have normally been trained on well-matched in-domain data. The adaptation of RNNLMs remains an open research area to be explored. In this paper, genre and topic based RNNLMadaptation techniques are investigated for a multi-genre broadcast transcription task. A number of techniques including Probabilistic Latent Semantic Analysis, Latent Dirichlet Allocation and Hierarchical Dirichlet Processes are used to extract show level topic information. These were then used as additional input to the RNNLM during training, which can facilitate unsupervised test time adaptation. Experiments using a state-of-theart LVCSR system trained on 1000 hours of speech and more than 1 billion words of text showed adaptation could yield perplexity reductions of 8% relatively over the baseline RNNLM and small but consistent word error rate reductions.

This paper presents the development of the 2014 Cambridge University conversational telephone Mandarin Chinese LVCSR system for the DARPA BOLT speech translation evaluation. A range of advanced modelling techniques were employed to both improve the recognition performance and provide a suitable integration with the translation system. These include an improved system combination technique using frame level acoustic model combination via joint decoding. Sequence trained deep neural network (DNN) based hybrid and tandem systems were combined on-the-fly to produce a consistent decoding output during search. A multi-level paraphrastic recurrent neural network LM (RNNLM) modelling both alternative paraphrase expressions and character sequences while preserving a consistent character to word segmentation was also used. This system gave an overall character error rate (CER) of 29.1% on the BOLT dev14 development set.

Recently a new approach to incorporate deep neural networks (DNN) bottleneck features into HMM based acoustic models using generalized variable parameter HMMs (GVPHMMs) was proposed. As Gaussian component level polynomial interpolation is performed for each high dimensional DNN bottleneck feature vector at a frame level, conventional GVPHMMs are computationally expensive to use in recognition time. To handle this problem, several approaches were exploited in this paper to efficiently use DNN bottleneck features in GVP-HMMs, including model selection techniques to optimally reduce the polynomial degrees; an efficient GMM based bottleneck feature clustering scheme; more compact GVP-HMM trajectory modelling for model space tied linear transformations. These improvements gave a total of 16 time speed up in decoding time over conventional GVP-HMMs using a uniformly assigned polynomial degree. Significant error rate reductions of 15.6% relative were obtained over the baseline tandem HMM system on the secondary microphone channel condition of Aurora 4 task. Consistent improvements were also obtained on other subsets.

Acoustic-to-articulatory inversion is useful for a range of related research areas including language learning, speech production, speech coding, speech recognition and speech synthesis. HMM-based generative modelling methods and DNNbased approaches have become dominant approaches in recent years. In this paper, a novel acoustic-to-articulatory inversion technique based on generalized variable parameter HMMs (GVP-HMMs) is proposed. It leverages the strengths of both generative and neural network based modelling frameworks. On a Mandarin speech inversion task, a tandem GVP-HMM system using DNN bottleneck features as auxiliary inputs significantly outperformed the baseline HMM, multiple regression HMM (MR-HMM), DNN and deep mixture density network (MDN) systems by 0.20mm, 0.16mm, 0.12mm and 0.10mm respectively in terms of electromagnetic articulography (EMA) root mean square error (RMSE).

An important part of the acoustic modelling problem for automatic speech recognition (ASR) systems is to handle the mismatch against a target environment created by time-varying external factors such as ambient noise. One possible solution to this problem is to introduce controllability to the underlying acoustic model to allow an instantaneous adaptation to the underlying noise condition. Along this line, the continuous trajectory of optimal, well matched model parameters against the varying noise can be explicitly modelled using, for example, generalized variable parameter HMMs (GVP-HMM). In order to improve the generalization and computational efficiency of conventional GVP-HMMs, this paper investigates a novel model complexity control method for GVP-HMMs. The optimal polynomial degrees of Gaussian mean, variance and model space linear transform trajectories are automatically determined at local level. Significant error rate reductions of 20% and 28% relative were obtained over the multi-style training baseline systems on Aurora 2 and a medium vocabulary Mandarin Chinese speech recognition task respectively. Consistent performance improvements and model size compression of 60% relative were also obtained over the baseline GVP-HMM systems using a uniformly assigned polynomial degree.

Expressive richness in natural languages presents a significant challenge for statistical language models (LM). As multiple word sequences can represent the same underlying meaning, only modelling the observed surface word sequence can lead to poor context coverage. To handle this issue, paraphrastic LMs were previously proposed to improve the generalization of back-off n-gram LMs. Paraphrastic neural network LMs (NNLM) are investigated in this paper. Using a paraphrastic multi-level feedforward NNLM modelling both word and phrase sequences, significant error rate reductions of 1.3% absolute (8% relative) and 0.9% absolute (5.5% relative) were obtained over the baseline n-gram and NNLM systems respectively on a state-of-the-art conversational telephone speech recognition system trained on 2000 hours of audio and 545 million words of texts.

Recurrent neural network language models (RNNLM) have become an increasingly popular choice for state-of-the-art speech recognition systems due to their inherently strong generalization performance. As these models use a vector representation of complete history contexts, RNNLMs are normally used to rescore N-best lists. Motivated by their intrinsic characteristics, two novel lattice rescoring methods for RNNLMs are investigated in this paper. The first uses an n-gram style clustering of history contexts. The second approach directly exploits the distance measure between hidden history vectors. Both methods produced 1-best performance comparable with a 10k-best rescoring baseline RNNLM system on a large vocabulary conversational telephone speech recognition task. Significant lattice size compression of over 70% and consistent improvements after confusion network (CN) decoding were also obtained over the N-best rescoring approach.

Recurrent neural network language models (RNNLMs) are be-coming increasingly popular for a range of applications includ-ing speech recognition. However, an important issue that limits the quantity of data, and hence their possible application ar-eas, is the computational cost in training. A standard approach to handle this problem is to use class-based outputs, allowing systems to be trained on CPUs. This paper describes an alter-native approach that allows RNNLMs to be efficiently trained on GPUs. This enables larger quantities of data to be used, and networks with an unclustered, full output layer to be trained. To improve efficiency on GPUs, multiple sentences are “spliced” together for each mini-batch or “bunch ” in training. On a large vocabulary conversational telephone speech recognition task, the training time was reduced by a factor of 27 over the stan-dard CPU-based RNNLM toolkit. The use of an unclustered, full output layer also improves perplexity and recognition per-formance over class-based RNNLMs. Index Terms: language models, recurrent neural network, speech recognition, GPU

Recently deep neural networks (DNNs) have become increasingly popular for acoustic modelling in automatic speech recognition (ASR) systems. As the bottleneck features they produce are inherently discriminative and contain rich hidden factors that influence the surface acoustic realization, the standard approach is to augment the conventional acoustic features with the bottleneck features in a tandem framework. In this paper, an alternative approach to incorporate bottleneck features is investigated. The complex relationship between acoustic features and DNN bottleneck features is modelled using generalized variable parameter HMMs (GVP-HMMs). The optimal GVP-HMM structural configuration and model parameters are automatically learnt. Significant error rate reductions of 48% and 8% relative were obtained over the baseline multi-style HMM and tandem HMM systems respectively on Aurora 2.

An important task for speech recognition systems is to handle the mismatch against a target environment introduced by acoustic factors such as variable ambient noise. To address this issue, it is possible to explicitly approximate the continuous trajectory of optimal, well matched model parameters against the varying noise using, for example, using generalized variable parameter HMMs (GVP-HMM). In order to improve the generalization and computational efficiency of conventional GVP-HMMs, this paper investigates a novel model complexity control method for GVP-HMMs. The optimal polynomial degrees of Gaussian mean, variance and model space linear transform trajectories are automatically determined at local level. Significant error rate reductions of 20% and 28% relative were obtained over the multi-style training baseline systems on Aurora 2 and a medium vocabulary Mandarin Chinese speech recognition task respectively. Consistent performance improvements and model size compression of 57% relative were also obtained over the baseline GVP-HMM systems using a uniformly assigned polynomial degree.

In natural languages multiple word sequences can represent the same underlying meaning. Only modelling the observed surface word sequence can result in poor context coverage, for example, when using n-gram language models (LM). To handle this issue, paraphrastic LMs were proposed in previous research and successfully applied to a US English conversational telephone speech transcription task. In order to exploit the complementary characteristics of paraphrastic LMs and neural network LMs (NNLM), the combination between the two is investigated in this paper. To investigate paraphrastic LMs' generalization ability to other languages, experiments are conducted on a Mandarin Chinese broadcast speech transcription task. Using a paraphrastic multi-level LM modelling both word and phrase sequences, significant error rate reductions of 0.9% absolute (9% relative) and 0.5% absolute (5% relative) were obtained over the baseline n-gram and NNLM systems respectively, after a combination with word and phrase level NNLMs.

Handling variable ambient noise is a challenging task for automatic speech recognition (ASR) systems. To address this issue, multi-style training using speech data collected in diverse noise environments, noise adaptive training or uncertainty decoding techniques can be used. An alternative approach is to explicitly approximate the continuous trajectory of Gaussian component or model space linear transform parameters against the varying noise, for example, using generalized variable parameter HMMs (GVP-HMM). In order to reduce the computational cost of conventional GVP-HMMs when model parameter update against the varying noise condition is required, this paper investigates a novel and more efficient extension of GVPHMMs that can also model the trajectories of feature space linear transforms. Significant error rate reductions of 9.3% and 18.5% relative were obtained over the multi-style training baseline system on Aurora 2 and a medium vocabulary Mandarin Chinese speech recognition task respectively.

In natural languages the variability in the underlying linguistic generation rules significantly alters the observed surface word sequence they create, and thus introduces a mismatch against other data generated via alternative realizations associated with, for example, a different domain. Hence, direct modelling of out-of-domain data can result in poor generalization to the indomain data of interest. To handle this problem, this paper investigated using cross-domain paraphrastic language models to improve in-domain language modelling (LM) using out-ofdomain data. Phrase level paraphrase models learnt from each domain were used to generate paraphrase variants for the data of other domains. These were used to both improve the context coverage of in-domain data, and reduce the domain mismatch of the out-of-domain data. Significant error rate reduction of 0.6% absolute was obtained on a state-of-the-art conversational telephone speech recognition task using a cross-domain paraphrastic multi-level LM trained on a billion words of mixed conversational and broadcast news data. Consistent improvements on the in-domain data context coverage were also obtained. Copyright © 2013 ISCA.

This paper investigates improving lightly supervised acoustic model training for an archive of broadcast data. Standard lightly supervised training uses automatically derived decoding hypotheses using a biased language model. However, as the actual speech can deviate significantly from the original programme scripts that are supplied, the quality of standard lightly supervised hypotheses can be poor. To address this issue, word and segment level combination approaches are used between the lightly supervised transcripts and the original programme scripts which yield improved transcriptions. Experimental results show that systems trained using these improved transcriptions consistently outperform those trained using only the original lightly supervised decoding hypotheses. This is shown to be the case for both the maximum likelihood and minimum phone error trained systems.

Natural languages are known for their expressive richness.Many sentences can be used to represent the same underlying meaning.Only modelling the observed surface word sequence can result in poor context coverage and generalization, for example, when using n-gram language models (LMs).This paper proposes a novel form of language model, the paraphrastic LM, that addresses these issues.A phrase level paraphrase model statistically learned from standard text data with no semantic annotation is used to generate multiple paraphrase variants.LM probabilities are then estimated by maximizing their marginal probability.Multi-level language models estimated at both the word level and the phrase level are combined.An efficient weighted finite state transducer (WFST) based paraphrase generation approach is also presented.Significant error rate reductions of 0.5%-0.6%absolute were obtained over the baseline n-gram LMs on two state-of-the-art recognition tasks for English conversational telephone speech and Mandarin Chinese broadcast speech using a paraphrastic multi-level LM modelling both word and phrase sequences.When it is further combined with word and phrase level feed-forward neural network LMs, a significant error rate reduction of 0.9% absolute (9% relative) and 0.5% absolute (5% relative) were obtained over the baseline n-gram and neural network LMs respectively.

We describe our work on developing a speech recognition system for multi-genre media archives. The high diversity of the data makes this a challenging recognition task, which may benefit from systems trained on a combination of in-domain and out-of-domain data. Working with tandem HMMs, we present Multi-level Adaptive Networks (MLAN), a novel technique for incorporating information from out-of-domain posterior features using deep neural networks. We show that it provides a substantial reduction in WER over other systems, with relative WER reductions of 15% over a PLP baseline, 9% over in-domain tandem features and 8% over the best out-of-domain tandem features.

One important issue in designing state-of-the-art LVCSR systems is the choice of acoustic units. Context dependent (CD) phones remain die dominant form of acoustic units. They can capture the co-articulatory effect in speech via explicit modelling. However, for other more complicated phonological processes, they rely on the implicit modelling ability of the underlying statistical models. Alternatively, it is possible to construct acoustic models based on higher level linguistic units, for example, syllables, to explicitly capture these complex patterns. When sufficient training data is available, this approach may show an advantage over implicit acoustic modelling. In this paper a wide range of acoustic units are investigated to improve LVCSR system performance. Significant error rate gains up to 7.1% relative (0.8% abs.) were obtained on a state-of-the-art Mandarin Chinese broadcast audio recognition task using word and syllable position dependent triphone and quinphone models.

Handling variable ambient noise is a challenging task for au-tomatic speech recognition (ASR) systems. To address this is-sue, multi-style, noise condition independent (CI) model train-ing using speech data collected in diverse noise environments, or uncertainty decoding techniques can be used. An alternative approach is to explicitly approximate the continuous trajectory of Gaussian component mean and variance parameters against the varying noise level, for example, using variable parameter HMMs (VP-HMM). This paper investigates a more generalized form of variable parameter HMMs (GVP-HMM). In addition to Gaussian component means and variances, it can also provide a more compact trajectory modelling for tied linear transforma-tions. An alternative noise condition dependent (CD) training algorithm is also proposed to handle the bias to training noise condition distribution. Consistent error rate gains were obtained over conventional VP-HMM mean and variance only trajectory modelling on a medium vocabulary Mandarin Chinese in-car navigation command recognition task. Index Terms: variable noise, generalized variable parameter HMM, noise robust speech recognition

This paper describes recent improvements to the Cambridge Arabic Large Vocabulary Continuous Speech Recognition (LVCSR) Speech-to-Text (STT) system. It is shown that word-boundary context markers provide a powerful method to enhance graphemic systems by implicit phonetic information, improving the modelling capability of graphemic systems. In addition, a robust technique for full covariance Gaussian modelling in the Minimum Phone Error (MPE) training framework is introduced. This reduces the full covariance training to a diagonal covariance training problem, thereby solving related robustness problems. The full system results show that the combined use of these and other techniques within a multi-branch combination framework reduces the Word Error Rate (WER) of the complete system by up to 5.9 % relative.

State-of-the-art large vocabulary continuous speech recognition (LVCSR) systems often combine outputs from multiple sub-systems developed at different sites. Cross system adaptation can be used as an alternative to direct hypothesis level combina-tion schemes such as ROVER. The standard approach involves only cross adapting acoustic models. To fully exploit the com-plimentary features among sub-systems, language model (LM) cross adaptation techniques can be used. Previous research on multi-level n-gram LM cross adaptation is extended to further include the cross adaptation of neural network LMs in this pa-per. Using this improved LM cross adaptation framework, sig-nificant error rate gains of 4.0%-7.1 % relative were obtained over acoustic model only cross adaptation when combining a range of Chinese LVCSR sub-systems used in the 2010 and 2011 DARPA GALE evaluations. 1.

In speech recognition systems language model (LMs) are often constructed by training and combining multiple n-gram models. They can be either used to represent different genres or tasks found in diverse text sources, or capture stochastic properties of different linguistic symbol sequences, for example, syllables and words. Unsupervised LM adaptation may also be used to further improve robustness to varying styles or tasks. When using these techniques, extensive software changes are often required. In this paper an alternative and more general approach based on weighted finite state transducers (WFSTs) is investigated for LM combination and adaptation. As it is entirely based on well-defined WFST operations, minimum change to decoding tools is needed. A wide range of LM combination configurations can be flexibly supported. An efficient on-the-fly WFST decoding algorithm is also proposed. Significant error rate gains of 7.3% relative were obtained on a state-of-the-art broadcast audio recognition task using a history dependently adapted multi-level LM modelling both syllable and word sequences.

State-of-the-art large vocabulary continuous speech recognition (LVCSR) systems often combine outputs from multiple sub-systems that may even be developed at different sites. Cross system adaptation, in which model adaptation is performed using the outputs from another sub-system, can be used as an alternative to hypothesis level combination schemes such as ROVER. Normally cross adaptation is only performed on the acoustic models. However, there are many other levels in LVCSR systems' modelling hierarchy where complimentary features may be exploited, for example, the sub-word and the word level, to further improve cross adaptation based system combination. It is thus interesting to also cross adapt language models (LMs) to capture these additional useful features. In this paper cross adaptation is applied to three forms of language models, a multi-level LM that models both syllable and word sequences, a word level neural network LM, and the linear combination of the two. Significant error rate reductions of 4.0-7.1% relative were obtained over ROVER and acoustic model only cross adaptation when combining a range of Chinese LVCSR sub-systems used in the 2010 and 2011 DARPA GALE evaluations.

Neural network language models (NNLM) have become an increasingly popular choice for large vocabulary continuous speech recognition (LVCSR) tasks, due to their inherent gener-alisation and discriminative power. This paper present two tech-niques to improve performance of standard NNLMs. First, the form of NNLM is modelled by introduction an additional out-put layer node to model the probability mass of out-of-shortlist (OOS) words. An associated probability normalisation scheme is explicitly derived. Second, a novel NNLM adaptation method using a cascaded network is proposed. Consistent WER reduc-tions were obtained on a state-of-the-art Arabic LVCSR task over conventional NNLMs. Further performance gains were also observed after NNLM adaptation.

The Chinese language is based on characters which are syllabic in nature. Since languages have syllabotactic rules which govern the construction of syllables and their allowed sequences, Chinese character sequence models can be used as a first level approximation of allowed syllable sequences. N-gram character sequence models were trained on 4.3 billion characters. Characters are used as a first level recognition unit with multiple pronunciations per character. For comparison the CU-HTK Mandarin word based system was used to recognize words which were then converted to character sequences. The character only system error rates for one best recognition were slightly worse than word based character recognition. However combining the two systems using log-linear combination gives better results than either system separately. An equally weighted combination gave consistent CER gains of 0.1-0.2% absolute over the word based standard system. Copyright © 2009 ISCA.

Language models (LMs) are often constructed by building multiple individual component models that are combined using context independent interpolation weights. By tuning these weights, using either perplexity or discriminative approaches, it is possible to adapt LMs to a particular task. This paper investigates the use of context dependent weighting in both interpolation and test-time adaptation of language models. Depending on the previous word contexts, a discrete history weighting function is used to adjust the contribution from each component model. As this dramatically increases the number of parameters to estimate, robust weight estimation schemes are required. Several approaches are described in this paper. The first approach is based on MAP estimation where interpolation weights of lower order contexts are used as smoothing priors. The second approach uses training data to ensure robust estimation of LM interpolation weights. This can also serve as a smoothing prior for MAP adaptation. A normalized perplexity metric is proposed to handle the bias of the standard perplexity criterion to corpus size. A range of schemes to combine weight information obtained from training data and test data hypotheses are also proposed to improve robustness during context dependent LM adaptation. In addition, a minimum Bayes' risk (MBR) based discriminative training scheme is also proposed. An efficient weighted finite state transducer (WFST) decoding algorithm for context dependent interpolation is also presented. The proposed technique was evaluated using a state-of-the-art Mandarin Chinese broadcast speech transcription task. Character error rate (CER) reductions up to 7.3 relative were obtained as well as consistent perplexity improvements. © 2012 Elsevier Ltd. All rights reserved.

Language models (LMs) are often constructed by building multiple component LMs that are combined using interpolation weights. By tuning these interpolation weights, using either perplexity or discriminative approaches, it is possible to adapt LMs to a particular task. In this work, improved LM adaptation is achieved by introducing context dependent interpolation weights. An important part of this new approach is obtaining robust estimation. Two schemes for this are described. The first is based on MAP estimation, where either global interpolation weights are used as priors, or context dependent interpolation priors obtained from the training data. The second scheme uses class based contexts to determine the interpolation weights. Both schemes are evaluated using unsupervised LM adaptation on a Mandarin broadcast transcription task. Consistent gains in perplexity using context dependent, rather than global, weights are observed as well as reductions in character error rate. 1.

This paper investigates unsupervised test-time adaptation of language models (LM) using discriminative methods for a Mandarin broadcast speech transcription and translation task. A standard approach to adapt interpolated language models to is to optimize the component weights byminimizing the perplexity on supervision data. This is a widely made approximation for language modeling in automatic speech recognition (ASR) systems. For speech translation tasks, it is unclear whether a strong correlation still exists between perplexity and various forms of error cost functions in recognition and translation stages. The proposed minimum Bayes risk (MBR) based approach provides a flexible framework for unsupervised LM adaptation. It generalizes to a variety of forms of recognition and translation error metrics. LM adaptation is performed at the audio document level using either the character error rate (CER), or translation edit rate (TER) as the cost function. An efficient parameter estimation scheme using the extended Baum-Welch (EBW) algorithm is proposed. Experimental results on a state-of-the-art speech recognition and translation system are presented. The MBR adapted language models gave the best recognition and translation performance and reduced the TER score by up to 0.54% absolute.

The majority of state-of-the-art speech recognition systems make use of system combination. The combination approaches adopted have traditionally been tuned to minimising word error rates (WERs). In recent years there has been a growing interest in taking the output from speech recognition systems in one language and translating it into another. This paper investigates the use of cross-site combination approaches in terms of both WER and impact on translation performance. In addition, the stages involved in modifying the output from a speech-to-text (STT) system to be suitable for translation are described. Two source languages, Mandarin and Arabic, are recognised and then translated using a phrase-based statistical machine translation system into English. Performance of individual systems and cross-site combination using cross-adaptation and ROVER are given. Results show that the best STT combination scheme in terms of WER is not necessarily the most appropriate when translating speech.

This paper describes the development of the CU-HTK Mandarin speech-to-text (STT) system and assesses its performance as part of a transcription-translation pipeline which converts broadcast Mandarin audio into English text. Recent improvements to the STT system are described and these give character error rate (CER) gains of 14.3% absolute for a broadcast conversation (BC) task and 5.1% absolute for a broadcast news (BN) task. The output of these STT systems is then post-processed, so that it consists of sentence-like segments, and translated into English text using a statistical machine translation (SMT) system. The performance of the transcription-translation pipeline is evaluated using the translation edit rate (TER) and BLEU metrics. It is shown that improving both the STT system and the post-STT segmentations can lower the TER scores by up to 5.3% absolute and increase the BLEU scores by up to 2.7% absolute.

Selecting the model structure with the "appropriate" complexity is a standard problem for training large-vocabulary continuous-speech recognition (LVCSR) systems. State-of-the-art LVCSR systems are highly complex. A wide variety of techniques may be used which alter the system complexity and word error rate (WER). Explicitly evaluating systems for all possible configurations is infeasible; hence, an automatic model complexity control criterion is highly desirable. Most existing complexity control schemes can be classified into two types, Bayesian learning techniques and information theory approaches. An implicit assumption is made in both, that increasing the likelihood on held-out data decreases the WER. However, this correlation has been found quite weak for current speech recognition systems. This paper presents a novel discriminative complexity control technique, the marginalization of a discriminative growth function. This is a closer approximation to the true WER than standard approaches. Experimental results on a standard LVCSR Switchboard task showed that marginalized discriminative growth functions outperforms manually tuned systems and conventional complexity control techniques, such as Bayesian information criterion (BIC), in terms of WER

This paper discusses the development of the CU-HTK Mandarin broadcast news (BN) transcription system. The Mandarin BN task includes a significant amount of English data. Hence techniques have been investigated to allow the same system to handle both Mandarin and English by augmenting the Mandarin training sets with English acoustic and language model training data. A range of acoustic models were built including models based on Gaussianised features, speaker adaptive training and feature-space MPE. A multi-branch system architecture is described in which multiple acoustic model types, alternate phone sets and segmentations can be used in a system combination framework to generate the final output. The final system shows state-of-the-art performance over a range of test sets

The paper details all aspects of the CUHTK 2004 Mandarin conversational telephone speech transcription system, but concentrates on the development of the acoustic models. As there are significant differences between the available training corpora, both in terms of topics of conversation and accents, forms of data normalisation and adaptive training techniques are investigated. The baseline discriminatively trained acoustic models are compared to a system built with a Gaussianisation front-end, a speaker adaptively trained system and an adaptively trained structured precision matrix system. The models are finally evaluated within a multi-pass, multi-branch, system combination framework.

The paper describes the use of several advanced acoustic modeling techniques for the 2004 CU-HTK large vocabulary speech recognition systems. These techniques include Gaussianization for speaker normalization, discriminative cluster adaptive training (CAT), subspace for precision and mean (SPAM) modeling of inverse covariances, and discriminative complexity control. Acoustic models featuring these techniques were integrated into a state-of-the-art 10 real-time multi-pass system with sophisticated adaptation for performance evaluation. Experimental results are presented on both broadcast news (BN) and conversational telephone speech (CTS) transcription tasks.

This paper discusses the Cambridge University HTK (CU-HTK) system for the automatic transcription of conversational telephone speech. A detailed discussion of the most important techniques in front-end processing, acoustic modeling and model training, language and pronunciation modeling are presented. These include the use of conversation side based cepstral normalization, vocal tract length normalization, heteroscedastic linear discriminant analysis for feature projection, minimum phone error training and speaker adaptive training, lattice-based model adaptation, confusion network based decoding and confidence score estimation, pronunciation selection, language model interpolation, and class based language models. The transcription system developed for participation in the 2002 NIST Rich Transcription evaluations of English conversational telephone speech data is presented in detail. In this evaluation the CU-HTK system gave an overall word error rate of 23.9%, which was the best performance by a statistically significant margin. Further details on the derivation of faster systems with moderate performance degradation are discussed in the context of the 2002 CU-HTK 10 /spl times/ RT conversational speech transcription system.

The paper describes the development of the 2003 CU-HTK large vocabulary speech recognition system for conversational telephone speech (CTS). The system was designed based on a multipass, multibranch structure where the output of all branches is combined using system combination. A number of advanced modelling techniques, such as speaker adaptive training, heteroscedastic linear discriminant analysis, minimum phone error estimation and specially constructed single pronunciation dictionaries, were employed. The effectiveness of each of these techniques and their potential contribution to the result of system combination was evaluated in the framework of a state-of-the-art LVCSR system with sophisticated adaptation. The final 2003 CU-HTK CTS system constructed from some of these models is described and its performance on the DARPA/NIST 2003 rich transcription (RT-03) evaluation test set is discussed.

State-of-the-art large vocabulary speech recognition systems are highly complex. Many techniques affect both system complexity and recognition performance. The need to determine the appropriate complexity without having to build each possible system has led to the development of automatic complexity control criteria. In this paper further experiments are carried out using a recently proposed criterion based on marginalizing a maximum mutual information (MMI) growth function. The use of this criterion is much detailed for determining the appropriate dimensionality in a multiple HLDA system and the number of components per state. A scheme for also using this criterion for model compression is described. Experimental results on a spontaneous telephone speech recognition task are described. Initial system compression experiments are inconclusive. However, comparing a standard state-of-the-art system with one generated using complexity control shows a reduction in word error rate.

Designing a state-of-the-art large vocabulary speech recognition systems is a highly complex problem. A wide range of techniques are available that affect the performance and number of free parameters. Selecting the appropriate complexity of system is both time-consuming and only a limited number of possible systems can be examined. This paper presents initial results on automatic system selection when both the number of dimensions and the number of components vary. Various complexity control schemes are discussed and evaluated. Limitations of schemes based on predicting held-out data log-likelihoods are described. In addition, problems of standard approximations for this task are detailed.