While large language models (LLMs) have been applied to automatic speech recognition (ASR), the task of making the model streamable remains a challenge. This paper proposes a novel model architecture, Transducer-Llama, that integrates LLMs into a Factorized Transducer (FT) model, naturally enabling streaming capabilities. Furthermore, given that the large vocabulary of LLMs can cause data sparsity issue and increased training costs for spoken language systems, this paper introduces an efficient vocabulary adaptation technique to align LLMs with speech system vocabularies. The results show that directly optimizing the FT model with a strong pre-trained LLM-based predictor using the RNN-T loss yields some but limited improvements over a smaller pre-trained LM predictor. Therefore, this paper proposes a weak-to-strong LM swap strategy, using a weak LM predictor during RNN-T loss training and then replacing it with a strong LLM. After LM replacement, the minimum word error rate (MWER) loss is employed to finetune the integration of the LLM predictor with the Transducer-Llama model. Experiments on the LibriSpeech and large-scale multi-lingual LibriSpeech corpora show that the proposed streaming Transducer-Llama approach gave a 17% relative WER reduction (WERR) over a strong FT baseline and a 32% WERR over an RNN-T baseline.

Modern smart glasses leverage machine learning to offer real-time transcriptions, considerably enriching human communication experiences. However, such systems frequently encounter challenges related to environmental noises, leading to decreased speech recognition. To improve voice quality, this work investigates directional source separation using the multi-microphone array. We explore multiple beamformers to assist source separation by strengthening the directional properties of speech signals. In addition to relying on predetermined beamformers, we investigate neural beamforming in multi-channel source separation, demonstrating that automatic learning directional characteristics effectively improves separation quality. Furthermore, we investigate the training strategies for ASR when utilizing separated outputs. Our results suggest that jointly training a directional speech separation and ASR model achieves the best overall performance while balancing the wearer and conversation partner’s performance.

We propose the joint speech translation and recognition (JSTAR) model that leverages the fast-slow cascaded encoder architecture for simultaneous end-to-end automatic speech recognition (ASR) and speech translation (ST). The model is transducer-based and uses a multi-objective training strategy that optimizes both ASR and ST objectives simultaneously. This allows JSTAR to produce high-quality streaming ASR and ST results. We apply JSTAR in a bilingual conversational speech setting with smart-glasses, where the model is also trained to distinguish speech from different directions corresponding to the wearer and a conversational partner. Different model pre-training strategies are studied to further improve results, including training of a transducer-based streaming machine translation (MT) model for the first time and applying it for parameter initialization of JSTAR. We demonstrate superior performances of JSTAR compared to a strong cascaded ST model in both BLEU scores and latency.

The growing popularity of multi-channel wearable devices, such as smart glasses, has led to a surge of applications such as targeted speech recognition and enhanced hearing. However, current approaches to solve these tasks use independently trained models, which may not benefit from large amounts of unlabeled data. In this paper, we propose M-BEST-RQ, the first multi-channel speech foundation model for smart glasses, which is designed to leverage large-scale self-supervised learning (SSL) in an array-geometry agnostic approach. While prior work on multi-channel speech SSL only evaluated on simulated settings, we curate a suite of real downstream tasks to evaluate our model, namely (i) conversational automatic speech recognition (ASR), (ii) spherical active source localization, and (iii) glasses wearer voice activity detection, which are sourced from the MMCSG and EasyCom datasets. We show that a general-purpose M-BEST-RQ encoder is able to match or surpass supervised models across all tasks. For the conversational ASR task in particular, using only 8 hours of labeled speech, our model outperforms a supervised ASR baseline that is trained on 2000 hours of labeled data, which demonstrates the effectiveness of our approach.

Cascaded speech-to-speech translation systems often suffer from the error accumulation problem and high latency, which is a result of cascaded modules whose inference delays accumulate. In this paper, we propose a transducer-based speech translation model that outputs discrete speech tokens in a low-latency streaming fashion. This approach eliminates the need for generating text output first, followed by machine translation (MT) and text-to-speech (TTS) systems. The produced speech tokens can be directly used to generate a speech signal with low latency by utilizing an acoustic language model (LM) to obtain acoustic tokens and an audio codec model to retrieve the waveform. Experimental results show that the proposed method outperforms other existing approaches and achieves state-of-the-art results for streaming translation in terms of BLEU, average latency, and BLASER 2.0 scores for multiple language pairs using the CVSS-C dataset as a benchmark.

The internal language model (ILM) of the neural transducer has been widely studied. In most prior work, it is mainly used for estimating the ILM score and is subsequently subtracted during inference to facilitate improved integration with external language models. Recently, various of factorized transducer models have been proposed, which explicitly embrace a standalone internal language model for non-blank token prediction. However, even with the adoption of factorized transducer models, limited improvement has been observed compared to shallow fusion. In this paper, we propose a novel ILM training and decoding strategy for factorized transducer models, which effectively combines the blank, acoustic and ILM scores. Our experiments show a 17% relative improvement over the standard decoding method when utilizing a well-trained ILM and the proposed decoding strategy on LibriSpeech datasets. Furthermore, when compared to a strong RNN-T baseline enhanced with external LM fusion, the proposed model yields a 5.5% relative improvement on general-sets and an 8.9% WER reduction for rare words. The proposed model can achieve superior performance without relying on external language models, rendering it highly efficient for production use-cases. To further improve the performance, we propose a novel and memory-efficient ILM-fusion-aware minimum word error rate (MWER) training method which improves ILM integration significantly.

Wearable devices like smart glasses are approaching the compute capability to seamlessly generate real-time closed captions for live conversations. We build on our recently introduced directional Automatic Speech Recognition (ASR) for smart glasses that have microphone arrays, which fuses multi-channel ASR with serialized output training, for wearer/conversation-partner disambiguation as well as suppression of cross-talk speech from non-target directions and noise.When ASR work is part of a broader system-development process, one may be faced with changes to microphone geometries as system development progresses.This paper aims to make multi-channel ASR insensitive to limited variations of microphone-array geometry. We show that a model trained on multiple similar geometries is largely agnostic and generalizes well to new geometries, as long as they are not too different. Furthermore, training the model this way improves accuracy for seen geometries by 15 to 28% relative. Lastly, we refine the beamforming by a novel Non-Linearly Constrained Minimum Variance criterion.

Recently reported state-of-the-art results in visual speech recognition (VSR) often rely on increasingly large amounts of video data, while the publicly available tran-scribed video datasets are limited in size. In this paper, for the first time, we study the potential of leveraging synthetic visual data for VSR. Our method, termed Synth VSR, sub-stantially improves the performance of VSR systems with synthetic lip movements. The key idea behind Synth VSR is to leverage a speech-driven lip animation model that gen-erates lip movements conditioned on the input speech. The speech-driven lip animation model is trained on an unla-beled audio-visual dataset and could be further optimized towards a pre-trained VSR model when labeled videos are available. As plenty of transcribed acoustic data and face images are available, we are able to generate large-scale synthetic data using the proposed lip animation model for semi-supervised VSR training. We evaluate the performance of our approach on the largest public VSR bench-mark - Lip Reading Sentences 3 (LRS3). Synth VSR achieves a WER of 43.3% with only 30 hours of real labeled data, outperforming off-the-shelf approaches using thousands of hours of video. The WER is further reduced to 27.9% when using all 438 hours of labeled data from LRS3, which is on par with the state-of-the-art self-supervised AV-HuBERT method. Furthermore, when combined with large-scale pseudo-labeled audio-visual data SynthVSR yields a new state-of-the-art VSR WER of 16.9% using publicly available data only, surpassing the recent state-of-the-art approaches trained with 29 times more non-public machine-transcribed video data (90,000 hours). Finally, we perform extensive ablation studies to understand the effect of each component in our proposed method.

Neural transducers have achieved human level performance on standard speech recognition benchmarks. However, their performance significantly degrades in the presence of cross-talk, especially when the primary speaker has a low signal-to-noise ratio. Anchored speech recognition refers to a class of methods that use information from an anchor segment (e.g., wake-words) to recognize device-directed speech while ignoring interfering background speech. In this paper, we investigate anchored speech recognition to make neural transducers robust to background speech. We extract context information from the anchor segment with a tiny auxiliary network, and use encoder biasing and joiner gating to guide the transducer towards the target speech. Moreover, to improve the robustness of context embedding extraction, we propose auxiliary training objectives to disentangle lexical content from speaking style. We evaluate our methods on synthetic LibriSpeech-based mixtures comprising several SNR and overlap conditions; they improve relative word error rates by 19.6% over a strong baseline, when averaged over all conditions.

Graph-based temporal classification (GTC), a generalized form of the connectionist temporal classification loss, was recently proposed to improve automatic speech recognition (ASR) systems using graph-based supervision. For example, GTC was first used to encode an N-best list of pseudo-label sequences into a graph for semi-supervised learning. In this paper, we propose an extension of GTC to model the posteriors of both labels and label transitions by a neural network, which can be applied to a wider range of tasks. As an example application, we use the extended GTC (GTC-e) for the multi-speaker speech recognition task. The transcriptions and speaker information of multi-speaker speech are represented by a graph, where the speaker information is associated with the transitions and ASR outputs with the nodes. Using GTC-e, multi-speaker ASR modelling becomes very similar to single-speaker ASR modeling, in that tokens by multiple speakers are recognized as a single merged sequence in chronological order. For evaluation, we perform experiments on a simulated multi-speaker speech dataset derived from LibriSpeech, obtaining promising results with performance close to classical benchmarks for the task.

Pseudo-labeling (PL), a semi-supervised learning (SSL) method where a seed model performs self-training using pseudo-labels generated from untranscribed speech, has been shown to enhance the performance of end-to-end automatic speech recognition (ASR). Our prior work proposed momentum pseudo-labeling (MPL), which performs PL-based SSL via an interaction between online and offline models, inspired by the mean teacher framework. MPL achieves remarkable results on various semi-supervised settings, showing robustness to variations in the amount of data and domain mismatch severity. However, there is further room for improving the seed model used to initialize the MPL training, as it is in general critical for a PL-based method to start training from high-quality pseudo-labels. To this end, we propose to enhance MPL by (1) introducing the Conformer architecture to boost the overall recognition accuracy and (2) exploiting iterative pseudo-labeling with a language model to improve the seed model before applying MPL. The experimental results demonstrate that the proposed approaches effectively improve MPL performance, outperforming other PL-based methods. We also present in-depth investigations to make our improvements effective, e.g., with regard to batch normalization typically used in Conformer and LM quality.

The recurrent neural network transducer (RNN-T) objective plays a major role in building today’s best automatic speech recognition (ASR) systems for production. Similarly to the connectionist temporal classification (CTC) objective, the RNN-T loss uses specific rules that define how a set of alignments is generated to form a lattice for the full-sum training. However, it is yet largely unknown if these rules are optimal and do lead to the best possible ASR results. In this work, we present a new transducer objective function that generalizes the RNN-T loss to accept a graph representation of the labels, thus providing a flexible and efficient framework to manipulate training lattices, e.g., for studying different transition rules, implementing different transducer losses, or restricting alignments. We demonstrate that transducer-based ASR with CTC-like lattice achieves better results compared to standard RNN-T, while also ensuring a strictly monotonic alignment, which will allow better optimization of the decoding procedure. For example, the proposed CTC-like transducer achieves an improvement of 4.8% on the test-other condition of LibriSpeech relative to an equivalent RNN-T based system.

The two most popular loss functions for streaming end-to-end automatic speech recognition (ASR) are RNN-Transducer (RNN-T) and connectionist temporal classification (CTC). Between these two loss types we can classify the monotonic RNN-T (MonoRNN-T) and the recently proposed CTC-like Transducer (CTC-T). Monotonic transducers have a few advantages. First, RNN-T can suffer from runaway hallucination, where a model keeps emitting non-blank symbols without advancing in time. Secondly, monotonic transducers consume exactly one model score per time step and are therefore more compatible with traditional FST-based ASR decoders. However, the MonoRNN-T so far has been found to have worse accuracy than RNN-T. It does not have to be that way: By regularizing the training via joint LAS training or parameter initialization from RNN-T, both MonoRNN-T and CTC-T perform as well or better than RNN-T. This is demonstrated for LibriSpeech and for a large-scale in-house data set.

The performance of automatic speech recognition (ASR) systems typically degrades significantly when the training and test data domains are mismatched. In this paper, we show that self-training (ST) combined with an uncertainty-based pseudo-label filtering approach can be effectively used for domain adaptation. We propose DUST, a dropout-based uncertainty-driven self-training technique which uses agreement between multiple predictions of an ASR system obtained for different dropout settings to measure the model’s uncertainty about its prediction. DUST excludes pseudo-labeled data with high uncertainties from the training, which leads to substantially improved ASR results compared to ST without filtering, and accelerates the training time due to a reduced training data set. Domain adaptation experiments using WSJ as a source domain and TED-LIUM 3 as well as SWITCHBOARD as the target domains show that up to 80% of the performance of a system trained on ground-truth data can be recovered.

Self-attention has become an important and widely used neural network component that helped to establish new state-of-the-art results for various applications, such as machine translation and automatic speech recognition (ASR). However, the computational complexity of self-attention grows quadratically with the input sequence length. This can be particularly problematic for applications such as ASR, where an input sequence generated from an utterance can be relatively long. In this work, we propose a combination of restricted self-attention and a dilation mechanism, which we refer to as dilated self-attention. The restricted self-attention allows attention to neighboring frames of the query at a high resolution, and the dilation mechanism summarizes distant information to allow attending to it with a lower resolution. Different methods for summarizing distant frames are studied, such as subsampling, mean-pooling, and attention-based pooling. ASR results demonstrate substantial improvements compared to restricted self-attention alone, achieving similar results compared to full-sequence based self-attention with a fraction of the computational costs.

Semi-supervised learning has demonstrated promising results in automatic speech recognition (ASR) by self-training using a seed ASR model with pseudo-labels generated for unlabeled data. The effectiveness of this approach largely relies on the pseudo-label accuracy, for which typically only the 1-best ASR hypothesis is used. However, alternative ASR hypotheses of an N-best list can provide more accurate labels for an unlabeled speech utterance and also reflect uncertainties of the seed ASR model. In this paper, we propose a generalized form of the connectionist temporal classification (CTC) objective that accepts a graph representation of the training labels. The newly proposed graph-based temporal classification (GTC) objective is applied for self-training with WFST-based supervision, which is generated from an N-best list of pseudo-labels. In this setup, GTC is used to learn not only a temporal alignment, similarly to CTC, but also a label alignment to obtain the optimal pseudo-label sequence from the weighted graph. Results show that this approach can effectively exploit an N-best list of pseudo-labels with associated scores, considerably outperforming standard pseudo-labeling, with ASR results approaching an oracle experiment in which the best hypotheses of the N-best lists are selected manually.

Pseudo-labeling (PL) has been shown to be effective in semisupervised automatic speech recognition (ASR), where a base model is self-trained with pseudo-labels generated from unlabeled data.While PL can be further improved by iteratively updating pseudo-labels as the model evolves, most of the previous approaches involve inefficient retraining of the model or intricate control of the label update.We present momentum pseudo-labeling (MPL), a simple yet effective strategy for semisupervised ASR.MPL consists of a pair of online and offline models that interact and learn from each other, inspired by the mean teacher method.The online model is trained to predict pseudo-labels generated on the fly by the offline model.The offline model maintains a momentum-based moving average of the online model.MPL is performed in a single training process and the interaction between the two models effectively helps them reinforce each other to improve the ASR performance.We apply MPL to an end-to-end ASR model based on the connectionist temporal classification.The experimental results demonstrate that MPL effectively improves over the base model and is scalable to different semi-supervised scenarios with varying amounts of data or domain mismatch.

This paper addresses end-to-end automatic speech recognition (ASR) for long audio recordings such as lecture and conversational speeches.Most end-to-end ASR models are designed to recognize independent utterances, but contextual information (e.g., speaker or topic) over multiple utterances is known to be useful for ASR.In our prior work, we proposed a contextexpanded Transformer that accepts multiple consecutive utterances at the same time and predicts an output sequence for the last utterance, achieving 5-15% relative error reduction from utterance-based baselines in lecture and conversational ASR benchmarks.Although the results have shown remarkable performance gain, there is still potential to further improve the model architecture and the decoding process.In this paper, we extend our prior work by (1) introducing the Conformer architecture to further improve the accuracy, (2) accelerating the decoding process with a novel activation recycling technique, and (3) enabling streaming decoding with triggered attention.We demonstrate that the extended Transformer provides state-of-the-art end-to-end ASR performance, obtaining a 17.3% character error rate for the HKUST dataset and 12.0%/6.3%word error rates for the Switchboard-300 Eval2000 CallHome/Switchboard test sets.The new decoding method reduces decoding time by more than 50% and further enables streaming ASR with limited accuracy degradation.

Attention-based end-to-end automatic speech recognition (ASR) systems have recently demonstrated state-of-the-art results for numerous tasks. However, the application of self-attention and attention-based encoder-decoder models remains challenging for streaming ASR, where each word must be recognized shortly after it was spoken. In this work, we present the dual causal/non-causal self-attention (DCN) architecture, which in contrast to restricted self-attention prevents the overall context to grow beyond the look-ahead of a single layer when used in a deep architecture. DCN is compared to chunk-based and restricted self-attention using streaming transformer and conformer architectures, showing improved ASR performance over restricted self-attention and competitive ASR results compared to chunk-based self-attention, while providing the advantage of frame-synchronous processing. Combined with triggered attention, the proposed streaming end-to-end ASR systems obtained state-of-the-art results on the LibriSpeech, HKUST, and Switchboard ASR tasks.

Encoder-decoder based sequence-to-sequence models have demonstrated state-of-the-art results in end-to-end automatic speech recognition (ASR). Recently, the transformer architecture, which uses self-attention to model temporal context information, has been shown to achieve significantly lower word error rates (WERs) compared to recurrent neural network (RNN) based system architectures. Despite its success, the practical usage is limited to offline ASR tasks, since encoder-decoder architectures typically require an entire speech utterance as input. In this work, we propose a transformer based end-to-end ASR system for streaming ASR, where an output must be generated shortly after each spoken word. To achieve this, we apply time-restricted self-attention for the encoder and triggered attention for the encoder-decoder attention mechanism. Our proposed streaming transformer architecture achieves 2.8% and 7.3% WER for the "clean" and "other" test data of LibriSpeech, which to our knowledge is the best published streaming end-to-end ASR result for this task.

We propose an unsupervised speaker adaptation method inspired by the neural Turing machine for end-to-end (E2E) automatic speech recognition (ASR). The proposed model contains a memory block that holds speaker i-vectors extracted from the training data and reads relevant i-vectors from the memory through an attention mechanism. The resulting memory vector (M-vector) is concatenated to the acoustic features or to the hidden layer activations of an E2E neural network model. The E2E ASR system is based on the joint connectionist temporal classification and attention-based encoder-decoder architecture. M-vector and i-vector results are compared for inserting them at different layers of the encoder neural network using the WSJ and TED-LIUM2 ASR benchmarks. We show that M-vectors, which do not require an auxiliary speaker embedding extraction system at test time, achieve similar word error rates (WERs) compared to i-vectors for single speaker utterances and significantly lower WERs for utterances in which there are speaker changes.

In this paper, we present a one-pass decoding algorithm for streaming recognition with joint connectionist temporal classification (CTC) and attention-based end-to-end automatic speech recognition (ASR) models. The decoding scheme is based on a frame-synchronous CTC prefix beam search algorithm and the recently proposed triggered attention concept. To achieve a fully streaming end-to-end ASR system, the CTC-triggered attention decoder is combined with a unidirectional encoder neural network based on parallel time-delayed long short-term memory (PTDLSTM) streams, which has demonstrated superior performance compared to various other streaming encoder architectures in earlier work. A new type of pre-training method is studied to further improve our streaming ASR models by adding residual connections to the encoder neural network and layer-wise removing them during the training process. The proposed joint CTC-triggered attention decoding algorithm, which enables streaming recognition of attention-based ASR systems, achieves similar ASR results compared to offline CTC-attention decoding and significantly better results compared to CTC prefix beam search decoding alone.

A new system architecture for end-to-end automatic speech recognition (ASR) is proposed that combines the alignment capabilities of the connectionist temporal classification (CTC) approach and the modeling strength of the attention mechanism. The proposed system architecture, named triggered attention (TA), uses a CTC-based classifier to control the activation of an attention-based decoder neural network. This allows for a frame-synchronous decoding scheme with an adjustable look-ahead parameter to control the induced delay and opens the door to streaming recognition with attention-based end-to-end ASR systems. We present ASR results of the TA model on three data sets of different size and language and compare the scores to a well-tuned attention-based end-to-end ASR baseline system, which consumes input frames in the traditional full-sequence manner. The proposed triggered attention (TA) decoder concept achieves similar or better ASR results in all experiments compared to the full-sequence attention model, while also limiting the decoding delay to two look-ahead frames, which in our setup corresponds to an output delay of 80 ms.

This paper evaluates neural network (NN) based systems and compares them to Gaussian mixture model (GMM) and hidden Markov model (HMM) approaches for acoustic scene classification (SC) and polyphonic acoustic event detection (AED) that are applied to data of the “Detection and Classification of Acoustic Scenes and Events 2016” (DCASE'16) challenge, task 1 and task 3, respectively. For both tasks, the use of deep neural networks (DNNs) and features based on an amplitude modulation filterbank and a Gabor filterbank (GFB) are evaluated and compared to standard approaches. For SC, additionally a time-delay NN approach is proposed that enables analysis of long contextual information similar to recurrent NNs but with training efforts comparable to conventional DNNs. The SC system proposed for task 1 of the DCASE'16 challenge attains a recognition accuracy of 77.5%, which is 5.6% higher compared to the DCASE'16 baseline system. For the AED task, DNNs are adopted in tandem and hybrid approaches, i.e., as part of HMM-based systems. These systems are evaluated for the polyphonic data of task 3 from the DCASE'16 challenge. Several strategies to address the issue of polyphony are considered. It is shown that DNN-based systems perform less accurate than the traditional systems for this task. Best results are achieved using GFB features in combination with a multiclass GMM-HMM back end.

S.2562-2566

Inspired by physiological studies on the human auditory system and by results from psychoacoustics, an amplitude modulation filter bank (AMFB) has been developed and successfully applied to feature extraction for automatic speech recognition (ASR) in earlier work. Here, we address the question as to which amplitude modulation (AM) frequency decomposition leads to optimal ASR performance by proposing a parameterized functional relationship between modulation center frequency and modulation bandwidth. Word error rates (WERs) of ASR experiments with 1551 different AMFBs are systematically evaluated and compared, resulting in the identification of a comparatively narrow range of optimal modulation frequency to modulation bandwidth characteristics. To integrate modulation processing with deep neural network (DNN) acoustic modeling, we propose merging of modulation filter coefficients with DNN weights prior to a final training step and an improved mean-variance normalization scheme for AMFBs. These modifications are shown to result in further reduction of WERs and are indicative of the proposed system's improved generalization ability, when compared across corpora of 100-960 h of data with mismatched training and test conditions. Analysis of DNN-learned temporal AM filtering properties is carried out and implications for the relevance of different modulation regions as well as the relation to psychoacoustic findings are discussed. ASR experiments with the proposed system demonstrate a high degree of robustness against extrinsic acoustic distortions, resulting in, e.g., an average WER of 9.79% on the Aurora-4 task.

The paper describes an automatic speech recognition (ASR) system for the 3rd CHiME challenge that addresses noisy acoustic scenes within public environments. The proposed system includes a multi-channel speech enhancement front-end including a microphone channel failure detection method that is based on cross-comparing the modulation spectra of speech to detect erroneous microphone recordings. The main focus of the submission is the investigation of the amplitude modulation filter bank (AMFB) as a method to extract long-term acoustic cues prior to a Gaussian mixture model (GMM) or deep neural network (DNN) based ASR classifier. It is shown that AMFB features outperform the commonly used frame splicing technique of filter bank features even on a performance optimized ASR challenge system. I.e., temporal analysis of speech by hand-crafted and auditory motivated AMFBs is shown to be more robust compared to a data-driven method based on extracting temporal dynamics with a DNN. Our final ASR system, which additionally includes adaptation of acoustic features to speaker characteristics, achieves an absolute word error rate reduction of approx. 21.53 % relative to the best CHiME-3 baseline system on the "real" test condition.

The human ability to classify acoustic sounds is still unmatched compared to recent methods in machine learning. Psychoacoustic and physiological studies indicate that the auditory system of mammals decomposes audio signals into their acoustic and modulation frequency components prior to further analysis. Since it is known that most linguistic information is coded in amplitude fluctuations, mimicking temporal processing strategies of the auditory system in automatic speech recognition (ASR) promises to increase recognition accuracies. We present an amplitude modulation filter bank (AMFB) that is used as a feature extraction scheme in ASR systems. The time-frequency resolution of the employed FIR filters, i.e., bandwidth and modulation frequency settings, are adopted from a psychophysically inspired model of Dau (1997) that was originally proposed to describe data from human psychoacoustics. Investigations on modulation phase indicate the need for preserving such information in amplitude modulation features. We show that the filter symmetry has an important impact on ASR performance. The proposed feature extraction scheme exhibits significant word error rate (WER) reductions using the Aurora-2, Aurora-4, and REVERB ASR tasks compared to other recent feature extraction methods, such as MFCC, FDLP, and PNCC features. Thereby, AMFB features reveal high robustness against additive noise, different transmission channel characteristics, and room reverberation. Using the Aurora-4 benchmark, for instance, an average WER of 12.33% with raw and 11.31% with bottleneck transformed features is attained, which constitutes a relative improvement of 19.6% and 29.2% over raw MFCC features, respectively.

Features inspired by the auditory system have previously demonstrated improvement in automatic speech recognition (ASR). Similarly, the use of Deep Neural Networks (DNN) was found to outperform classic approaches to ASR in many conditions. Since DNNs have the potential to learn the task relevant features from a conventional filter bank output, we investigate if the combination of auditory features and deep learning should be preferred over self-learned patterns. Specifically, noise-robust Gabor features and Amplitude Modulation Filter-Bank (AMFB) features, highly invariant against reverberation, are used as input to a state-of-the-art ASR system incorporating DNN processing. On the Aurora-4 task, both mel-frequency cepstral coefficients (MFCC) and filter bank (FBank) features are outperformed in many acoustic conditions through auditory processing, yielding average relative improvements of up to 69% over MFCC and 21% over the commonly used DNN-FBank setup. This highlights the mutual benefit of auditory signal processing and recent advances in machine learning.

Many research efforts in the field of feature extraction for automatic speech recognition are focused on analyzing slow amplitude fluctuations of speech. In this study the importance of spectral and temporal resolution for the amplitude modulation frequency analysis are investigated in order to provide guidance for the appropriate filter design. Therefore, different wavelet and Fourier transform like filter time scales are examined, i.e. The importance of time and frequency separation is compared. The results demonstrate that analyzing three separate amplitude modulation frequency bands of constant absolute bandwidth that cover the range from about 2 to 16 Hz are sufficient for automatic speech recognition.

In this contribution we present a feature extraction method that relies on the modulation-spectral analysis of amplitude fluctuations within sub-bands of the acoustic spectrum by a STFT. The experimental results indicate that the optimal temporal filter extension for amplitude modulation analysis is around 310 ms. It is also demonstrated that the phase information of the modulation spectrum contains important cues for speech recognition. In this context, the advantage of an odd analysis basis function is considered. The best presented features reached a total relative improvement of 53.5% for clean-condition training on Aurora-2. Furthermore, it is shown that modulation features are more robust against room reverberation than conventional cepstral and dynamic features and that they strongly benefit from a high early-to-late energy ratio of the characteristic RIR.