Several studies have shown promising results in adapting DNN-based acoustic models as a mechanism to transfer characteristics from pre-trained models. One such example is speaker adaptation using a small amount of data, where fine-tuning has helped train models that extrapolate well to diverse linguistic contexts that are not present in the adaptation data. In the current work, our objective is to synthesize speech in different languages using the target speaker's voice, regardless of the language of their data. To achieve this goal, we create a multilingual model using a corpus that consists of recordings from a large number of monolingual and a few bilingual speakers in multiple languages. The model is then adapted using the target speaker's recordings in a language other than the target language. We also explore if additional adaptation data from a native speaker of the target language improves the performance. The subjective evaluation shows that the proposed approach of cross-language speaker adaptation is able to synthesize speech in the target language, in the target speaker's voice, without data spoken by the target speaker in that language. Also, extra data from a native speaker of the target language can improve model performance.

We present in this work a strategy to enhance the quality of Text-to-Speech (TTS) based Singing Voice generation. Speech-to-singing refers to techniques transforming a spoken voice into singing, mainly by manipulating the duration and pitch of a spoken version of a song's lyrics. While this strategy efficiently preserves the speaker identity, the generated singing is not always perceived fully natural since the vocal conditions generally change between spoken and singing voice. By incorporating speaker-independent natural singing information to TTS-based Speech-to-Singing (STS) we positively impact the sound quality (e.g. reducing hoarseness), as it is shown in the subjective evaluation reported at the end of this paper.

This work investigates techniques that select training data from small, found corpuses in order to improve the naturalness of synthesized text-to-speech voices. The approach outlined in this paper examines different metrics to detect and reject segments of training data that can degrade the performance of the system. We conducted experiments on two small datasets extracted from Mandarin Chinese audiobooks that have different characteristics in terms of recording conditions, narrator, and transcriptions. We show that using a even smaller, yet carefully selected, set of data can lead to a text-to-speech system able to generate more natural speech than a system trained on the complete dataset. Three metrics related to the narrator's articulation proposed in the paper give significant improvements in naturalness.

Most of the degradation in current Statistical Parametric Speech Synthesis (SPSS) results from the form of the vocoder. One of the main causes of degradation is the reconstruction of the noise. In this article, a new signal model is proposed that leads to a simple synthesizer, without the need for ad-hoc tuning of model parameters. The model is not based on the traditional additive linear source-filter model, it adopts a combination of speech components that are additive in the log domain. Also, the same representation for voiced and unvoiced segments is used, rather than relying on binary voicing decisions. This avoids voicing error discontinuities that can occur in many current vocoders. A simple binary mask is used to denote the presence of noise in the time-frequency domain, which is less sensitive to classification errors. Four experiments have been carried out to evaluate this new model. The first experiment examines the noise reconstruction issue. Three listening tests have also been carried out that demonstrate the advantages of this model: comparison with the STRAIGHT vocoder; the direct prediction of the binary noise mask by using a mixed output configuration; and partial improvements of creakiness using a mask correction mechanism.

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.

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.

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) 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.

Personalisation of voice output communication aids (VOCAs) allows to preserve the vocal identity of people suffering from speech disorders. This can be achieved by the adaptation of HMM-based speech synthesis systems using a small amount of adaptation data. When the voice has begun to deteriorate, reconstruction is still possible in the statistical domain by correcting the parameters of the models associated with the speech disorder. This can be done by substituting those with parameters from a donor’s voice, at risk of losing part of the identity of the patient. Recently, the Multiple-Average-Voice-Model (Multiple AVM) framework has been proposed for speaker adaptation. Adaptation is performed via interpolation into a speaker eigenspace spanned by the mean vectors of speaker-adapted AVMs which can be tuned to the individual speaker. In this paper, we present the benefits of this framework for voice reconstruction: it requires only a very small amount of adaptation data, interpolation can be performed in a clean speech eigenspace and the resulting voice can be easily fine-tuned by acting on the interpolation weights. We illustrate our points with a subjective assessment of the reconstructed voice. Index Terms: HMM-Based speech synthesis, speaker adaptation, multiple average voice model, cluster adaptive training, voice reconstruction, voice output communication aids.

The assignment of prosodic events (accent and phrasing) from the text is crucial in text-to-speech synthesis systems. This paper addresses the combination of linguistic and metric constraints for the assignment of prosodic events in text-to-speech synthesis. First, a linguistic processing chain is used to provide a rich linguistic description of a text. Then, a novel statistical representation based on a hierarchical HMM (HHMM) is used to model the prosodic structure of a text: the root layer represents the text, each intermediate layer a sequence of intermediate phrases, the pre-terminal layer the sequence of accents, and the terminal layer the sequence of linguistic contexts. For each intermediate layer, a segmental HMM and information fusion are used to fuse the linguistic and metric constraints for the segmentation of a text into phrases. A set of experiments conducted on multi-speaker databases with various speaking styles reports that: the rich linguistic representation improves drastically the assignment of prosodic events, and the fusion of linguistic and metric constraints significantly improves over standard methods for the segmentation of a text into phrases. These constitute substantial advances that can be further used to model the speech prosody of a speaker, a speaking style, and emotions for text-to-speech synthesis.

This paper describes a novel approach for the speaker adaptation of statistical parametric speech synthesis systems based on the interpolation of a set of average voice models (AVM). Recent results have shown that the quality/naturalness of adapted voices depends on the distance from the average voice model used for speaker adaptation. This suggests the use of several AVMs trained on carefully chosen speaker clusters from which a more suitable AVM can be selected/interpolated during the adaptation. In the proposed approach a set of AVMs, a multiple-AVM, is trained on distinct clusters of speakers which are iteratively re-assigned during the estimation process initialised according to metadata. During adaptation, each AVM from the multiple-AVM is first adapted towards the target speaker. The adapted means from the AVMs are then interpolated to yield the final speaker adapted mean for synthesis. It is shown, performing speaker adaptation on a corpus of British speakers with various regional accents, that the quality/naturalness of synthetic speech of adapted voices is significantly higher than when considering a single factor-independent AVM selected according to the target speaker characteristics.

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.

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.

Spectral voice conversion is usually performed using a single model selected in order to represent a tradeoff between goodness of fit and complexity. Recently, we proposed a new method for spectral voice conversion, called Dynamic Model Selection (DMS), in which we assumed that the model topology may change over time, depending on the source acoustic features. In this method a set of models with increasing complexity is considered during the conversion of a source speech signal into a target speech signal. During the conversion, the best model is dynamically selected among the models in the set, according to the acoustical features of each source frame. In this paper, we present an objective evaluation demonstrating that this new method improves the conversion by reducing the transformation error compared to methods based on an single model.

In this paper, a method for prosodic break modelling based on segmental-HMMs and Dempster-Shafer fusion for speech synthesis is presented, and the relative importance of linguistic and metric constraints in prosodic break modelling is assessed 1 . A context-dependent segmental-HMM is used to explicitly model the linguistic and the metric constraints. Dempster-Shafer fusion is used to balance the relative importance of the linguistic and the metric constraints into the segmental-HMM. A linguistic processing chain based on surface and deep syntactic parsing is additionally used to extract linguistic informations of different nature. An objective evaluation proved evidence that the optimal combination of the linguistic and the metric constraints significantly outperforms both the conventional HMM (linguistic information only) and segmental-HMM (equal balance of linguistic and metric constraints), and confirmed that the linguistic constraint is prior to the metric. Index Terms: speech prosody, prosodic break, segmentalHMM, Dempster-Shafer fusion.

This paper assesses the ability of a HMM-based speech synthesis systems to model the speech characteristics of various speaking styles 1 .A discrete/continuous HMM is presented to model the symbolic and acoustic speech characteristics of a speaking style.The proposed model is used to model the average characteristics of a speaking style that is shared among various speakers, depending on specific situations of speech communication.The evaluation consists of an identification experiment of 4 speaking styles based on delexicalized speech, and compared to a similar experiment on natural speech.The comparison is discussed and reveals that discrete/continuous HMM consistently models the speech characteristics of a speaking style.

This paper introduces a HMM-based speech synthesis system which uses a new method for the Separation of Vocal-tract and Liljencrants-Fant model plus Noise (SVLN). The glottal source is separated into two components: a deterministic glottal waveform Liljencrants-Fant model and a modulated Gaussian noise. This glottal source is first estimated and then used in the vocal-tract estimation procedure. Then, the parameters of the source and the vocal-tract are included into HMM contextual models of phonems. SVLN is promising for voice transformation in synthesis of expressive speech since it allows an independent control of vocal-tract and glottal-source properties. The synthesis results are finally discussed and subjectively evaluated.

cote interne IRCAM: Lanchantin10b