Robust Speech Processing in Adverse Environments 1
HomeFull List of Titles
1: ICSLP'98 Proceedings
Keynote SpeechesText-To-Speech Synthesis 1Spoken Language Models and Dialog 1Prosody and Emotion 1Hidden Markov Model Techniques 1Speaker and Language Recognition 1Multimodal Spoken Language Processing 1Isolated Word RecognitionRobust Speech Processing in Adverse Environments 1Spoken Language Models and Dialog 2Articulatory Modelling 1Talking to Infants, Pets and LoversRobust Speech Processing in Adverse Environments 2Spoken Language Models and Dialog 3Speech Coding 1Articulatory Modelling 2Prosody and Emotion 2Neural Networks, Fuzzy and Evolutionary Methods 1Utterance Verification and Word Spotting 1 / Speaker Adaptation 1Text-To-Speech Synthesis 2Spoken Language Models and Dialog 4Human Speech Perception 1Robust Speech Processing in Adverse Environments 3Speech and Hearing Disorders 1Prosody and Emotion 3Spoken Language Understanding Systems 1Signal Processing and Speech Analysis 1Spoken Language Generation and Translation 1Spoken Language Models and Dialog 5Segmentation, Labelling and Speech Corpora 1Multimodal Spoken Language Processing 2Prosody and Emotion 4Neural Networks, Fuzzy and Evolutionary Methods 2Large Vocabulary Continuous Speech Recognition 1Speaker and Language Recognition 2Signal Processing and Speech Analysis 2Prosody and Emotion 5Robust Speech Processing in Adverse Environments 4Segmentation, Labelling and Speech Corpora 2Speech Technology Applications and Human-Machine Interface 1Large Vocabulary Continuous Speech Recognition 2Text-To-Speech Synthesis 3Language Acquisition 1Acoustic Phonetics 1Speaker Adaptation 2Speech Coding 2Hidden Markov Model Techniques 2Multilingual Perception and Recognition 1Large Vocabulary Continuous Speech Recognition 3Articulatory Modelling 3Language Acquisition 2Speaker and Language Recognition 3Text-To-Speech Synthesis 4Spoken Language Understanding Systems 4Human Speech Perception 2Large Vocabulary Continuous Speech Recognition 4Spoken Language Understanding Systems 2Signal Processing and Speech Analysis 3Human Speech Perception 3Speaker Adaptation 3Spoken Language Understanding Systems 3Multimodal Spoken Language Processing 3Acoustic Phonetics 2Large Vocabulary Continuous Speech Recognition 5Speech Coding 3Language Acquisition 3 / Multilingual Perception and Recognition 2Segmentation, Labelling and Speech Corpora 3Text-To-Speech Synthesis 5Spoken Language Generation and Translation 2Human Speech Perception 4Robust Speech Processing in Adverse Environments 5Text-To-Speech Synthesis 6Speech Technology Applications and Human-Machine Interface 2Prosody and Emotion 6Hidden Markov Model Techniques 3Speech and Hearing Disorders 2 / Speech Processing for the Speech and Hearing Impaired 1Human Speech ProductionSegmentation, Labelling and Speech Corpora 4Speaker and Language Recognition 4Speech Technology Applications and Human-Machine Interface 3Utterance Verification and Word Spotting 2Large Vocabulary Continuous Speech Recognition 6Neural Networks, Fuzzy and Evolutionary Methods 3Speech Processing for the Speech-Impaired and Hearing-Impaired 2Prosody and Emotion 72: SST Student Day
SST Student Day - Poster Session 1SST Student Day - Poster Session 2Author Index
A B C D E F G H I
J K L M N O P Q R
S T U V W X Y Z
Multimedia Files
Robust Speech Recognition using HMM's with Toeplitz State Covariance matrices
Authors:
William J.J. Roberts, Defence Science Technology Organisation (Australia)
Yariv Ephraim, George Mason University (USA)
Page (NA) Paper number 141
Abstract:
Hidden Markov modeling of speech waveforms is studied and applied to speech recognition of clean and noisy signals. Signal vectors in each state are assumed Gaussian with zero mean and a Toeplitz covariance matrix. This model allows short signal vectors and thus is useful for speech signals with rapidly changing second order statistics. It can also be straightforwardly adapted to noisy signals especially when the noise is additive and independent of the signal. Since no closed form solution exists for the maximum likelihood estimate of the Toeplitz covariance matrices, an expectation-maximization procedure was used and efficiently implemented. HMM's with Toeplitz as well as asymptotically Toeplitz (e.g., circulant, autoregressive) covariance matrices are theoretically and experimentally studied. While asymptotically all of these matrices provide similar performance, they differ significantly when the frame length is finite. Recognition results are provided for clean and noisy signals at 0-30dB SNR.
SL980141.PDF (From Author)
SL980141.PDF (Rasterized)
TOP
Modeling of Output Probability Distribution to Improve Small Vocabulary Speech Recognition in Adverse Environments
Authors:
David Thambiratnam, Queensland University of Technology (Australia)
Sridha Sridharan, Queensland University of Technology (Australia)
Page (NA) Paper number 308
Abstract:
This paper presents a solution to the adverse environment, open microphone problem, by using the information stored in HMM output probability distributions to obtain a confidence measure of the results. This information can also be used to perform a secondary classification and improve recognition results. The system was tested on data from the TI46 database that had been corrupted by noise from the NOISEX-92 database, as well as on real-world data, and shows promising results.
SL980308.PDF (From Author)
SL980308.PDF (Rasterized)
TOP
Robust and Compact Multilingual Word Recognizers Using Features Extracted from a Phoneme Similarity Front-End
Authors:
Philippe Morin, Panasonic Technologies, Inc. / Speech Technology Laboratory (USA)
Ted H. Applebaum, Panasonic Technologies, Inc. / Speech Technology Laboratory (USA)
Robert Boman, Panasonic Technologies, Inc. / Speech Technology Laboratory (USA)
Yi Zhao, Panasonic Technologies, Inc. / Speech Technology Laboratory (USA)
Jean-Claude Junqua, Panasonic Technologies, Inc. / Speech Technology Laboratory (USA)
Page (NA) Paper number 402
Abstract:
In this paper we characterize the sensitivity of two speaker-dependent isolated word recognizers toward several kinds of variability and distortions; namely noise, channels, distance to microphone and target language. Both recognizers use a phoneme similarity acoustic front-end as a rich representation for speech from which reliable features are extracted. A cross-correlation test showed that a phoneme similarity front-end is more robust to variability and distortions (especially intra-speaker variability) than a LPC cepstral front-end. The first recognizer (Condor) uses a frame-based approach while the second (Pasha) uses the phoneme similarity information contained in a small number of speech segments. The two recognition methods are presented with a special emphasis on the robustness improvements and computational trade-offs that have been made. Experimental results are reported for car noise at different speeds, speakerphone versus handset input in an office environment and several target languages. Recognition accuracy greater than 94% was achieved in a car environment at 60 mph (Condor) and recognition accuracy greater than 95% was achieved for speakerphone input at a distance of 50 cm. in an office environment.
SL980402.PDF (From Author)
SL980402.PDF (Rasterized)
TOP
An Effect of Adaptive Beamforming on Hands-Free Speech Recognition Based on 3-D Viterbi Search
Authors:
Takeshi Yamada, Graduate School of Information Science, Nara Institute of Science and Technology (Japan)
Satoshi Nakamura, Graduate School of Information Science, Nara Institute of Science and Technology (Japan)
Kiyohiro Shikano, Graduate School of Information Science, Nara Institute of Science and Technology (Japan)
Page (NA) Paper number 484
Abstract:
To integrate the microphone array processing into speech recognition, we have proposed a speech recognition algorithm based on 3-D Viterbi search, which localizes a target talker considering the likelihood of HMMs (Hidden Markov Models) while performing speech recognition. The performance of the 3-D Viterbi search method depends on the improvement of the SNR (Signal to Noise Ratio) by the beamforming technique. This paper proposes a novel method based on an adaptive beamforming technique instead of the delay-and-sum beamformer used in our previous study. The speaker-dependent isolated-word recognition experiments were carried out on real environment data to evaluate the effect of the adaptive beamformer. These results showed that the adaptive beamformer drastically improves the recognition performance both for a fixed-position talker and for a moving-talker.
SL980484.PDF (From Author)
SL980484.PDF (Rasterized)
TOP
Coherence-based Subband Decomposition for Robust Speech and Speaker Recognition in Noisy and Reverberant Rooms
Authors:
Joaquin González-Rodríguez, DIAC- Universidad Politecnica de Madrid (Spain)
Santiago Cruz-Llanas, DIAC- Universidad Politecnica de Madrid (Spain)
Javier Ortega-García, DIAC- Universidad Politecnica de Madrid (Spain)
Page (NA) Paper number 64
Abstract:
In this paper, the acoustic characteristics of sound fields in enclosed rooms are studied in the joint presence of speech and noise, in order to design a broadband microphone array system capable of coping with both coherent and diffuse noises. Several state-of-the-art speech enhancement array structures are presented and compared to our new system in terms of correct word recognition rates in a simple command and control task. The proposed structure, based on a broadband subband-nested array, performs real-time estimations of the spatial coherence in order to determine the coherent/diffuse nature of the different subbands, using different filters in each case, improving also the classical Wiener post-filter, typically used for diffuse noise supression, for proper cancellation of coherent noises. The results obtained with a 15-channel simultaneous recording database in different reverberation and noise conditions show better performance than other structures previously proposed.
SL980064.PDF (From Author)
SL980064.PDF (Rasterized)
TOP
A Minimax Search Algorithm for CDHMM based Robust Continuous Speech Recognition
Authors:
Hui Jiang, The University of Tokyo (Japan)
Keikichi Hirose, The University of Tokyo (Japan)
Qiang Huo, The University of Hong Kong (China)
Page (NA) Paper number 693
Abstract:
In this paper, we propose a novel implementation of a minimax decision rule for continuous density hidden Markov model based robust speech recognition. By combining the idea of the minimax decision rule with a normal Viterbi search, we derive a recursive minimax search algorithm, where the minimax decision rule is repetitively applied to determine the partial paths during the search procedure. Because of its intrinsic nature of a recursive search, the proposed method can be easily extended to perform continuous speech recognition. Experimental results on Japanese isolated digits and TIDIGITS, where the mismatch between training and testing conditions is caused by additive white Gaussian noise, show the viability and efficiency of the proposed minimax search algorithm.