Progress on Phoneme Recognition with a Continuous-State HMM

Phil Weber; Linxue Bai; Stephen Houghton; Peter Jancovic; Martin Russell

doi:10.1109/ICASSP.2016.7472799

Progress on Phoneme Recognition with a Continuous-State HMM

Phil Weber, Linxue Bai, Stephen Houghton, Peter Jancovic, Martin Russell

Electronic, Electrical and Systems Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

3 Citations (Scopus)

Abstract

Recent advances in automatic speech recognition have used large corpora and powerful computational resources to train complex statistical models from high-dimensional features, to attempt to capture all the variability found in natural speech. Such models are difficult to interpret and may be fragile, and contradict or ignore knowledge of human speech production and perception. We report progress towards phoneme recognition using a model of speech which employs very few parameters and which is more faithful to the dynamics and model of human speech production. Using features generated from a neural network bottleneck layer, we obtain recognition accuracy on TIMIT which compares favourably with traditional models of similar power. We discuss the implications of these results for recognition using natural features such as vocal tract resonances and spectral energies

Original language	English
Title of host publication	2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP): Proceedings
Publisher	IEEE Xplore
Number of pages	5
ISBN (Electronic)	2379-190X
DOIs	https://doi.org/10.1109/ICASSP.2016.7472799
Publication status	E-pub ahead of print - 19 May 2016
Event	IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP) 2016 - China, Shanghai, China Duration: 20 Mar 2016 → 25 Mar 2016 http://www.icassp2016.org

Conference

Conference	IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP) 2016
Country/Territory	China
City	Shanghai
Period	20/03/16 → 25/03/16
Internet address	http://www.icassp2016.org

Keywords

Hidden Markov models
speech
training
video recording
speech recognition
computational modeling
data models

Access to Document

10.1109/ICASSP.2016.7472799Licence: None: All rights reserved

http://ieeexplore.ieee.org/document/7472799/Licence: None: All rights reserved

Cite this

@inproceedings{29fde198f53c44e4a65844360d2cec0d,

title = "Progress on Phoneme Recognition with a Continuous-State HMM",

abstract = "Recent advances in automatic speech recognition have used large corpora and powerful computational resources to train complex statistical models from high-dimensional features, to attempt to capture all the variability found in natural speech. Such models are difficult to interpret and may be fragile, and contradict or ignore knowledge of human speech production and perception. We report progress towards phoneme recognition using a model of speech which employs very few parameters and which is more faithful to the dynamics and model of human speech production. Using features generated from a neural network bottleneck layer, we obtain recognition accuracy on TIMIT which compares favourably with traditional models of similar power. We discuss the implications of these results for recognition using natural features such as vocal tract resonances and spectral energies",

keywords = "Hidden Markov models , speech , training , video recording , speech recognition , computational modeling , data models",

author = "Phil Weber and Linxue Bai and Stephen Houghton and Peter Jancovic and Martin Russell",

year = "2016",

month = may,

day = "19",

doi = "10.1109/ICASSP.2016.7472799",

language = "English",

booktitle = "2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP): Proceedings",

publisher = "IEEE Xplore",

note = "IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP) 2016 ; Conference date: 20-03-2016 Through 25-03-2016",

url = "http://www.icassp2016.org",

}

Weber, P, Bai, L, Houghton, S, Jancovic, P & Russell, M 2016, Progress on Phoneme Recognition with a Continuous-State HMM. in 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP): Proceedings. IEEE Xplore, IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP) 2016, Shanghai, China, 20/03/16. https://doi.org/10.1109/ICASSP.2016.7472799

TY - GEN

T1 - Progress on Phoneme Recognition with a Continuous-State HMM

AU - Weber, Phil

AU - Bai, Linxue

AU - Houghton, Stephen

AU - Jancovic, Peter

AU - Russell, Martin

PY - 2016/5/19

Y1 - 2016/5/19

N2 - Recent advances in automatic speech recognition have used large corpora and powerful computational resources to train complex statistical models from high-dimensional features, to attempt to capture all the variability found in natural speech. Such models are difficult to interpret and may be fragile, and contradict or ignore knowledge of human speech production and perception. We report progress towards phoneme recognition using a model of speech which employs very few parameters and which is more faithful to the dynamics and model of human speech production. Using features generated from a neural network bottleneck layer, we obtain recognition accuracy on TIMIT which compares favourably with traditional models of similar power. We discuss the implications of these results for recognition using natural features such as vocal tract resonances and spectral energies

AB - Recent advances in automatic speech recognition have used large corpora and powerful computational resources to train complex statistical models from high-dimensional features, to attempt to capture all the variability found in natural speech. Such models are difficult to interpret and may be fragile, and contradict or ignore knowledge of human speech production and perception. We report progress towards phoneme recognition using a model of speech which employs very few parameters and which is more faithful to the dynamics and model of human speech production. Using features generated from a neural network bottleneck layer, we obtain recognition accuracy on TIMIT which compares favourably with traditional models of similar power. We discuss the implications of these results for recognition using natural features such as vocal tract resonances and spectral energies

KW - Hidden Markov models

KW - speech

KW - training

KW - video recording

KW - speech recognition

KW - computational modeling

KW - data models

U2 - 10.1109/ICASSP.2016.7472799

DO - 10.1109/ICASSP.2016.7472799

M3 - Conference contribution

BT - 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP): Proceedings

PB - IEEE Xplore

T2 - IEEE Int. Conf. on Acoustics, Speech and Signal Processing (ICASSP) 2016

Y2 - 20 March 2016 through 25 March 2016

ER -

Progress on Phoneme Recognition with a Continuous-State HMM

Abstract

Conference

Keywords

Access to Document

Fingerprint

Cite this