Sorry, you need to enable JavaScript to visit this website.

facebooktwittermailshare

FREQUENCY DOMAIN MULTI-CHANNEL ACOUSTIC MODELING FOR DISTANT SPEECH RECOGNITION

Abstract: 

Conventional far-field automatic speech recognition (ASR) systems typically employ microphone array techniques for speech enhancement in order to improve robustness against noise or reverberation. However, such speech enhancement techniques do not always yield ASR accuracy improvement because the optimization criterion for speech enhancement is not directly relevant to the ASR objective. In this work, we develop new acoustic modeling techniques that optimize spatial filtering and long short-term memory (LSTM) layers from multi-channel (MC) input based on an ASR criterion directly. In contrast to conventional methods, we incorporate array processing knowledge into the acoustic model. Moreover, we initialize the network with beamformers' coefficients. We investigate effects of such MC neural networks through ASR experiments on the real-world far-field data where users are interacting with an ASR system in uncontrolled acoustic environments. We show that our MC acoustic model can reduce a word error rate (WER) by 16.5% compared to a single channel ASR system with the traditional log-mel filter bank energy (LFBE) feature on average. Our result also shows that our network with the spatial filtering layer on two-channel input achieves a relative WER reduction of 9.5% compared to conventional beamforming with seven microphones.

up
0 users have voted:

Paper Details

Authors:
Shiva Sundaram, Nikko Strom, Bjorn Hoffmeister
Submitted On:
10 May 2019 - 6:36pm
Short Link:
Type:
Poster
Event:
Presenter's Name:
Kenichi Kumatani
Paper Code:
SLP-P15.4
Document Year:
2019
Cite

Document Files

poster file

(17)

manuscript file

(15)

Subscribe

[1] Shiva Sundaram, Nikko Strom, Bjorn Hoffmeister, "FREQUENCY DOMAIN MULTI-CHANNEL ACOUSTIC MODELING FOR DISTANT SPEECH RECOGNITION", IEEE SigPort, 2019. [Online]. Available: http://sigport.org/4419. Accessed: Jun. 25, 2019.
@article{4419-19,
url = {http://sigport.org/4419},
author = {Shiva Sundaram; Nikko Strom; Bjorn Hoffmeister },
publisher = {IEEE SigPort},
title = {FREQUENCY DOMAIN MULTI-CHANNEL ACOUSTIC MODELING FOR DISTANT SPEECH RECOGNITION},
year = {2019} }
TY - EJOUR
T1 - FREQUENCY DOMAIN MULTI-CHANNEL ACOUSTIC MODELING FOR DISTANT SPEECH RECOGNITION
AU - Shiva Sundaram; Nikko Strom; Bjorn Hoffmeister
PY - 2019
PB - IEEE SigPort
UR - http://sigport.org/4419
ER -
Shiva Sundaram, Nikko Strom, Bjorn Hoffmeister. (2019). FREQUENCY DOMAIN MULTI-CHANNEL ACOUSTIC MODELING FOR DISTANT SPEECH RECOGNITION. IEEE SigPort. http://sigport.org/4419
Shiva Sundaram, Nikko Strom, Bjorn Hoffmeister, 2019. FREQUENCY DOMAIN MULTI-CHANNEL ACOUSTIC MODELING FOR DISTANT SPEECH RECOGNITION. Available at: http://sigport.org/4419.
Shiva Sundaram, Nikko Strom, Bjorn Hoffmeister. (2019). "FREQUENCY DOMAIN MULTI-CHANNEL ACOUSTIC MODELING FOR DISTANT SPEECH RECOGNITION." Web.
1. Shiva Sundaram, Nikko Strom, Bjorn Hoffmeister. FREQUENCY DOMAIN MULTI-CHANNEL ACOUSTIC MODELING FOR DISTANT SPEECH RECOGNITION [Internet]. IEEE SigPort; 2019. Available from : http://sigport.org/4419