Jean-Luc Gauvain
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Jean-Luc Gauvain has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Lori Lamel
CNRS/LIMSI, Spoken Language Processing Group
3 shared publications - Claude Barras
LIMSI/CNRS, Université Paris-Sud
2 shared publications Lori Lamel
Spoken Language Processing Group, LIMSI-CNRS
2 shared publications- Alexandre Allauzen
LAMSADE - Laboratoire d'analyse et modélisation de systèmes pour l'aide à la décision
1 shared publication - Anindya Roy
LIMSI/CNRS
1 shared publication - Cheung Chi Leung
Human Language Technology Department, Institute for Infocomm Research (I2R), Connexis, Singapore
1 shared publication
Latest Publications
- 9 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 26 (3) , 646-656
Optimization of RNN-Based Speech Activity Detection
Speech activity detection (SAD) is an essential component of automatic speech recognition (ASR) systems impacting the overall system performance. This paper investigates an optimization process for recurrent neural network (RNN) based SAD. This process optimizes all system parameters including those used for feature extraction, the NN weights, and the back- end parameters. Three cost functions are considered for SAD optimization: the frame error rate (FER), the NIST detection cost function (DCF), and the word error rate (WER) of a downstream speech recognizer. Different types of RNN models and optimization methods are investigated. Three types of RNNs are compared: a basic RNN, an LSTM network with peepholes, and a coordinated-gate LSTM network introduced in [1]. Well suited for non-differentiable optimization problems, quantum-behaved particle swarm optimization (QPSO) is used to optimize feature extraction and posterior smoothing, as well as for the initial training of the neural networks. Experimental SAD results are reported on the NIST 2015 SAD evaluation data as well as REPERE and AMI meeting corpora. Speech recognition results are reported on the OpenKWS'13 test data. For all tasks and conditions, the proposed optimization method significantly improves the SAD performance and among all the tested SAD methods the CG-LSTM model gives the best results.
- 9 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Vol. 26 (3) , 646-656
Optimization of RNN-Based Speech Activity Detection
Speech activity detection (SAD) is an essential component of automatic speech recognition (ASR) systems impacting the overall system performance. This paper investigates an optimization process for recurrent neural network (RNN) based SAD. This process optimizes all system parameters including those used for feature extraction, the NN weights, and the back- end parameters. Three cost functions are considered for SAD optimization: the frame error rate (FER), the NIST detection cost function (DCF), and the word error rate (WER) of a downstream speech recognizer. Different types of RNN models and optimization methods are investigated. Three types of RNNs are compared: a basic RNN, an LSTM network with peepholes, and a coordinated-gate LSTM network introduced in [1]. Well suited for non-differentiable optimization problems, quantum-behaved particle swarm optimization (QPSO) is used to optimize feature extraction and posterior smoothing, as well as for the initial training of the neural networks. Experimental SAD results are reported on the NIST 2015 SAD evaluation data as well as REPERE and AMI meeting corpora. Speech recognition results are reported on the OpenKWS'13 test data. For all tasks and conditions, the proposed optimization method significantly improves the SAD performance and among all the tested SAD methods the CG-LSTM model gives the best results.
- 17 November 2015
- Published by Springer Nature
- p. 164-172
Conversational Telephone Speech Recognition for Lithuanian
- 17 November 2015
- Published by Springer Nature
- p. 164-172
Conversational Telephone Speech Recognition for Lithuanian
- 17 April 2014
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 74 (4) , 1377-1396
Lexical speaker identification in TV shows
- 17 April 2014
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 74 (4) , 1377-1396
Lexical speaker identification in TV shows
- 18 September 2012
- Published by Oxford University Press (OUP)
Speech Recognition
Speech recognition is concerned with converting the speech waveform, an acoustic signal, into a sequence of words. Today's approaches are based on a statistical modellization of the speech signal. This article provides an overview of the main topics addressed in speech recognition, which are, acoustic-phonetic modelling, lexical representation, language modelling, decoding, and model adaptation. Language models are used in speech recognition to estimate the probability of word sequences. The main components of a generic speech recognition system are, main knowledge sources, feature analysis, and acoustic and language models, which are estimated in a training phase, and the decoder. The focus of this article is on methods used in state-of-the-art speaker-independent, large-vocabulary continuous speech recognition (LVCSR). Primary application areas for such technology are dictation, spoken language dialogue, and transcription for information archival and retrieval systems. Finally, this article discusses issues and directions of future research.
- 18 September 2012
- Published by Oxford University Press (OUP)
Speech Recognition
Speech recognition is concerned with converting the speech waveform, an acoustic signal, into a sequence of words. Today's approaches are based on a statistical modellization of the speech signal. This article provides an overview of the main topics addressed in speech recognition, which are, acoustic-phonetic modelling, lexical representation, language modelling, decoding, and model adaptation. Language models are used in speech recognition to estimate the probability of word sequences. The main components of a generic speech recognition system are, main knowledge sources, feature analysis, and acoustic and language models, which are estimated in a training phase, and the decoder. The focus of this article is on methods used in state-of-the-art speaker-independent, large-vocabulary continuous speech recognition (LVCSR). Primary application areas for such technology are dictation, spoken language dialogue, and transcription for information archival and retrieval systems. Finally, this article discusses issues and directions of future research.
- 1 January 2008
- Published by Association for Computational Linguistics (ACL)
LIMSI's statistical translation systems for WMT'08
- 1 January 2008
- Published by Association for Computational Linguistics (ACL)
LIMSI's statistical translation systems for WMT'08
- 1 April 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4 (15206149) , IV-641-641
Modeling Duration via Lattice Rescoring
It is often acknowledged that HMMs do not properly model phone and word durations. In this paper phone and word duration models are used to improve the accuracy of state-of-the-art large vocabulary speech recognition systems. The duration information is integrated into the systems in a rescoring of word lattices that include phone-level segmentations. Experimental results are given for a conversational telephone speech (CTS) task in French and for the TC-Star EPPS transcription task in Spanish and English. An absolute word error rate reduction of about 0.5% is observed for the CTS task, and smaller but consistent gains are observed for the EPPS task in both languages.
- 1 April 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4 (15206149) , IV-641-641
Modeling Duration via Lattice Rescoring
It is often acknowledged that HMMs do not properly model phone and word durations. In this paper phone and word duration models are used to improve the accuracy of state-of-the-art large vocabulary speech recognition systems. The duration information is integrated into the systems in a rescoring of word lattices that include phone-level segmentations. Experimental results are given for a conversational telephone speech (CTS) task in French and for the TC-Star EPPS transcription task in Spanish and English. An absolute word error rate reduction of about 0.5% is observed for the CTS task, and smaller but consistent gains are observed for the EPPS task in both languages.
- 1 April 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, IV-53-IV-56
Constrained MLLR for Speaker Recognition
One particularly difficult challenge for cross-channel MLLR (CMLLR) are two widely-used techniques for speaker introduced in the 2005 and 2006 NIST Speaker Recognition Evaluations, where training uses telephone speech and verification uses speech from multiple auxiliary comparable to that obtained with cepstral features. This paper describes a new feature extraction technique for speaker recognition based on CMLLR speaker adaptation which session effects through latent factor analysis (LFA) and through support vector machines (SVM). Results on the NIST operates directly on the recorded signal with noise well as in combination with two cepstral approaches such as reduction in the performance gap between telephone and auxiliary microphone data.
- 1 April 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, IV-53-IV-56
Constrained MLLR for Speaker Recognition
One particularly difficult challenge for cross-channel MLLR (CMLLR) are two widely-used techniques for speaker introduced in the 2005 and 2006 NIST Speaker Recognition Evaluations, where training uses telephone speech and verification uses speech from multiple auxiliary comparable to that obtained with cepstral features. This paper describes a new feature extraction technique for speaker recognition based on CMLLR speaker adaptation which session effects through latent factor analysis (LFA) and through support vector machines (SVM). Results on the NIST operates directly on the recorded signal with noise well as in combination with two cepstral approaches such as reduction in the performance gap between telephone and auxiliary microphone data.
- Open Access
- 1 January 2005
- Published by Association for Computational Linguistics (ACL)
- p. 201-208
Training neural network language models on very large corpora
During the last years there has been growing interest in using neural networks for language modeling. In contrast to the well known back-off n-gram language models, the neural network approach attempts to overcome the data sparseness problem by performing the estimation in a continuous space. This type of language model was mostly used for tasks for which only a very limited amount of in-domain training data is available. In this paper we present new algorithms to train a neural network language model on very large text corpora. This makes possible the use of the approach in domains where several hundreds of millions words of texts are available. The neural network language model is evaluated in a state-ofthe-art real-time continuous speech recognizer for French Broadcast News. Word error reductions of 0.5% absolute are reported using only a very limited amount of additional processing time.
- 1 January 2005
- Published by Association for Computational Linguistics (ACL)
- p. 201-208
Open AccessTraining neural network language models on very large corpora
During the last years there has been growing interest in using neural networks for language modeling. In contrast to the well known back-off n-gram language models, the neural network approach attempts to overcome the data sparseness problem by performing the estimation in a continuous space. This type of language model was mostly used for tasks for which only a very limited amount of in-domain training data is available. In this paper we present new algorithms to train a neural network language model on very large text corpora. This makes possible the use of the approach in domains where several hundreds of millions words of texts are available. The neural network language model is evaluated in a state-ofthe-art real-time continuous speech recognizer for French Broadcast News. Word error reductions of 0.5% absolute are reported using only a very limited amount of additional processing time.
- Open Access
- 25 February 2003
- Published by Springer Nature in EURASIP Journal on Advances in Signal Processing
- Vol. 2003 (2) , 642019-150
Structuring Broadcast Audio for Information Access
- 25 February 2003
- Published by Springer Nature in EURASIP Journal on Advances in Signal Processing
- Vol. 2003 (2) , 642019-150
Open AccessStructuring Broadcast Audio for Information Access
- 1 June 2001
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 14 (2) , 187-200
Audio Partitioning and Transcription for Broadcast Data Indexation
- 1 June 2001
- Published by Springer Nature in Multimedia Tools and Applications
- Vol. 14 (2) , 187-200
Audio Partitioning and Transcription for Broadcast Data Indexation