Saved in:
Bibliographic Details
Main Authors: Manohar, Vimal, Chen, Szu-Jui, Wang, Zhiqi, Fujita, Yusuke, Watanabe, Shinji, Khudanpur, Sanjeev
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2405.11078
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909206780575744
author Manohar, Vimal
Chen, Szu-Jui
Wang, Zhiqi
Fujita, Yusuke
Watanabe, Shinji
Khudanpur, Sanjeev
author_facet Manohar, Vimal
Chen, Szu-Jui
Wang, Zhiqi
Fujita, Yusuke
Watanabe, Shinji
Khudanpur, Sanjeev
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2405_11078
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Acoustic modeling for Overlapping Speech Recognition: JHU Chime-5 Challenge System
Manohar, Vimal
Chen, Szu-Jui
Wang, Zhiqi
Fujita, Yusuke
Watanabe, Shinji
Khudanpur, Sanjeev
Audio and Speech Processing
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.
title Acoustic modeling for Overlapping Speech Recognition: JHU Chime-5 Challenge System
topic Audio and Speech Processing
url https://arxiv.org/abs/2405.11078