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

facebooktwittermailshare

ON DNN POSTERIOR PROBABILITY COMBINATION IN MULTI-STREAM SPEECH RECOGNITION FOR REVERBERANT ENVIRONMENTS

Abstract: 

A multi-stream framework with deep neural network (DNN) classifiers has been applied in this paper to improve automatic speech recognition (ASR) performance in environments with different reverberation characteristics. We propose a room parameter estimation model to determine the stream weights for DNN posterior probability combination with the aim of obtaining reliable log-likelihoods for decoding. The model is implemented by training a multi-layer
perceptron to distinguish between various reverberant environments. The method is tested in known and unknown environments against approaches based on inverse entropy and autoencoders, with average relative word error rate improvements of 46% and 29%, respectively, when performing multi-stream ASR in different reverberant situations.

up
1 user has voted: Feifei Xiong

Paper Details

Authors:
Feifei Xiong, Stefan Goetze, Bernd T. Meyer
Submitted On:
28 February 2017 - 2:30am
Short Link:
Type:
Poster
Event:
Presenter's Name:
Feifei Xiong
Paper Code:
SP-P4.9
Document Year:
2017
Cite

Document Files

poster_icassp17_xiongetal.pdf

(342)

Subscribe

[1] Feifei Xiong, Stefan Goetze, Bernd T. Meyer, "ON DNN POSTERIOR PROBABILITY COMBINATION IN MULTI-STREAM SPEECH RECOGNITION FOR REVERBERANT ENVIRONMENTS", IEEE SigPort, 2017. [Online]. Available: http://sigport.org/1482. Accessed: Jan. 23, 2020.
@article{1482-17,
url = {http://sigport.org/1482},
author = {Feifei Xiong; Stefan Goetze; Bernd T. Meyer },
publisher = {IEEE SigPort},
title = {ON DNN POSTERIOR PROBABILITY COMBINATION IN MULTI-STREAM SPEECH RECOGNITION FOR REVERBERANT ENVIRONMENTS},
year = {2017} }
TY - EJOUR
T1 - ON DNN POSTERIOR PROBABILITY COMBINATION IN MULTI-STREAM SPEECH RECOGNITION FOR REVERBERANT ENVIRONMENTS
AU - Feifei Xiong; Stefan Goetze; Bernd T. Meyer
PY - 2017
PB - IEEE SigPort
UR - http://sigport.org/1482
ER -
Feifei Xiong, Stefan Goetze, Bernd T. Meyer. (2017). ON DNN POSTERIOR PROBABILITY COMBINATION IN MULTI-STREAM SPEECH RECOGNITION FOR REVERBERANT ENVIRONMENTS. IEEE SigPort. http://sigport.org/1482
Feifei Xiong, Stefan Goetze, Bernd T. Meyer, 2017. ON DNN POSTERIOR PROBABILITY COMBINATION IN MULTI-STREAM SPEECH RECOGNITION FOR REVERBERANT ENVIRONMENTS. Available at: http://sigport.org/1482.
Feifei Xiong, Stefan Goetze, Bernd T. Meyer. (2017). "ON DNN POSTERIOR PROBABILITY COMBINATION IN MULTI-STREAM SPEECH RECOGNITION FOR REVERBERANT ENVIRONMENTS." Web.
1. Feifei Xiong, Stefan Goetze, Bernd T. Meyer. ON DNN POSTERIOR PROBABILITY COMBINATION IN MULTI-STREAM SPEECH RECOGNITION FOR REVERBERANT ENVIRONMENTS [Internet]. IEEE SigPort; 2017. Available from : http://sigport.org/1482