Summary
2007 International Symposium on Nonlinear Theory and its Applications
2007
Session Number:19AM2-C
Session:
Number:19AM2-C-4
A study on a speech recognition method based on the selective sound segregation in various noisy environments
Atsushi Haniu, Masahi Unoki, Masato Akagi,
pp.445-448
Publication Date:2007/9/16
Online ISSN:2188-5079
DOI:10.34385/proc.41.19AM2-C-4
Summary:
This paper shows the effectiveness of our proposed speech recognition method in the noisy environments. In our proposed method, the target sound is recognized based on verifying the possibility of the existence of the target sound by evaluating the selective sound segregation using the hypothesis of the target sound and Bregman’s regularities. Since this method does not use any noise model, the proposed method is able to recognize the target sound under various noisy environments. In order to evaluate the proposed method, performance of ASR system using the proposed method in noisy environments was compared with performance of traditional methods under speech recognitions. The results show that the proposed method is more robust than other methods in experimental conditions from SNR = 0 dB to ∞dB. Therefore, these indicate the proposed method is a useful speech recognition concept in noisy environments.