Acoustic Modeling for Overlapping Speech Recognition: Jhu Chime-5 Challenge System

Vimal Manohar; Szu Jui Chen; Zhiqi Wang; Y. Fujita; Shinji Watanabe; Sanjeev Khudanpur

doi:10.1109/ICASSP.2019.8682556

Acoustic Modeling for Overlapping Speech Recognition: Jhu Chime-5 Challenge System

Vimal Manohar, Szu Jui Chen, Zhiqi Wang, Y. Fujita, Shinji Watanabe, Sanjeev Khudanpur

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

2 Citations (Scopus)

Abstract

This paper summarizes our acoustic modeling efforts in the Johns Hopkins University speech recognition system for the CHiME-5 challenge to recognize highly-overlapped dinner party speech recorded by multiple microphone arrays. We explore data augmentation approaches, neural network architectures, front-end speech dereverberation, beamforming and robust i-vector extraction with comparisons of our in-house implementations and publicly available tools. We finally achieved a word error rate of 69.4% on the development set, which is a 11.7% absolute improvement over the previous baseline of 81.1%, and release this improved baseline with refined techniques/tools as an advanced CHiME-5 recipe.

Original language	English
Title of host publication	2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	6665-6669
Number of pages	5
ISBN (Electronic)	9781479981311
DOIs	https://doi.org/10.1109/ICASSP.2019.8682556
Publication status	Published - 2019 May 1
Externally published	Yes
Event	44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Brighton, United Kingdom Duration: 2019 May 12 → 2019 May 17

Publication series

Name	ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume	2019-May
ISSN (Print)	1520-6149

Conference

Conference	44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019
Country	United Kingdom
City	Brighton
Period	19/5/12 → 19/5/17

Keywords

acoustic modeling
CHiME-5 challenge
Kaldi
Robust speech recognition

ASJC Scopus subject areas

Software
Signal Processing
Electrical and Electronic Engineering

Access to Document

10.1109/ICASSP.2019.8682556

Fingerprint Dive into the research topics of 'Acoustic Modeling for Overlapping Speech Recognition: Jhu Chime-5 Challenge System'. Together they form a unique fingerprint.

Cite this

Manohar, V., Chen, S. J., Wang, Z., Fujita, Y., Watanabe, S., & Khudanpur, S. (2019). Acoustic Modeling for Overlapping Speech Recognition: Jhu Chime-5 Challenge System. In 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings (pp. 6665-6669). [8682556] (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2019-May). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICASSP.2019.8682556

Acoustic Modeling for Overlapping Speech Recognition : Jhu Chime-5 Challenge System. / Manohar, Vimal; Chen, Szu Jui; Wang, Zhiqi; Fujita, Y.; Watanabe, Shinji; Khudanpur, Sanjeev.

2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2019. p. 6665-6669 8682556 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings; Vol. 2019-May).

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

Manohar, V, Chen, SJ, Wang, Z, Fujita, Y, Watanabe, S & Khudanpur, S 2019, Acoustic Modeling for Overlapping Speech Recognition: Jhu Chime-5 Challenge System. in 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings., 8682556, ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, vol. 2019-May, Institute of Electrical and Electronics Engineers Inc., pp. 6665-6669, 44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019, Brighton, United Kingdom, 19/5/12. https://doi.org/10.1109/ICASSP.2019.8682556

Manohar V, Chen SJ, Wang Z, Fujita Y, Watanabe S, Khudanpur S. Acoustic Modeling for Overlapping Speech Recognition: Jhu Chime-5 Challenge System. In 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2019. p. 6665-6669. 8682556. (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings). https://doi.org/10.1109/ICASSP.2019.8682556

Manohar, Vimal ; Chen, Szu Jui ; Wang, Zhiqi ; Fujita, Y. ; Watanabe, Shinji ; Khudanpur, Sanjeev. / Acoustic Modeling for Overlapping Speech Recognition : Jhu Chime-5 Challenge System. 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2019. pp. 6665-6669 (ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings).

@inproceedings{713e7ce29b94489e834379ea7d7eba95,

title = "Acoustic Modeling for Overlapping Speech Recognition: Jhu Chime-5 Challenge System",

abstract = "This paper summarizes our acoustic modeling efforts in the Johns Hopkins University speech recognition system for the CHiME-5 challenge to recognize highly-overlapped dinner party speech recorded by multiple microphone arrays. We explore data augmentation approaches, neural network architectures, front-end speech dereverberation, beamforming and robust i-vector extraction with comparisons of our in-house implementations and publicly available tools. We finally achieved a word error rate of 69.4% on the development set, which is a 11.7% absolute improvement over the previous baseline of 81.1%, and release this improved baseline with refined techniques/tools as an advanced CHiME-5 recipe.",

keywords = "acoustic modeling, CHiME-5 challenge, Kaldi, Robust speech recognition",

author = "Vimal Manohar and Chen, {Szu Jui} and Zhiqi Wang and Y. Fujita and Shinji Watanabe and Sanjeev Khudanpur",

year = "2019",

month = may,

day = "1",

doi = "10.1109/ICASSP.2019.8682556",

language = "English",

series = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "6665--6669",

booktitle = "2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings",

note = "44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 ; Conference date: 12-05-2019 Through 17-05-2019",

}

TY - GEN

T1 - Acoustic Modeling for Overlapping Speech Recognition

T2 - 44th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019

AU - Manohar, Vimal

AU - Chen, Szu Jui

AU - Wang, Zhiqi

AU - Fujita, Y.

AU - Watanabe, Shinji

AU - Khudanpur, Sanjeev

PY - 2019/5/1

Y1 - 2019/5/1

N2 - This paper summarizes our acoustic modeling efforts in the Johns Hopkins University speech recognition system for the CHiME-5 challenge to recognize highly-overlapped dinner party speech recorded by multiple microphone arrays. We explore data augmentation approaches, neural network architectures, front-end speech dereverberation, beamforming and robust i-vector extraction with comparisons of our in-house implementations and publicly available tools. We finally achieved a word error rate of 69.4% on the development set, which is a 11.7% absolute improvement over the previous baseline of 81.1%, and release this improved baseline with refined techniques/tools as an advanced CHiME-5 recipe.

AB - This paper summarizes our acoustic modeling efforts in the Johns Hopkins University speech recognition system for the CHiME-5 challenge to recognize highly-overlapped dinner party speech recorded by multiple microphone arrays. We explore data augmentation approaches, neural network architectures, front-end speech dereverberation, beamforming and robust i-vector extraction with comparisons of our in-house implementations and publicly available tools. We finally achieved a word error rate of 69.4% on the development set, which is a 11.7% absolute improvement over the previous baseline of 81.1%, and release this improved baseline with refined techniques/tools as an advanced CHiME-5 recipe.

KW - acoustic modeling

KW - CHiME-5 challenge

KW - Kaldi

KW - Robust speech recognition

UR - http://www.scopus.com/inward/record.url?scp=85068973715&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068973715&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2019.8682556

DO - 10.1109/ICASSP.2019.8682556

M3 - Conference contribution

AN - SCOPUS:85068973715

T3 - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

SP - 6665

EP - 6669

BT - 2019 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2019 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 12 May 2019 through 17 May 2019

ER -