Improving Speech Enhancement Using Soft-Constrained Inverse Filter in Noisy Reverberant Conditions

Ken'ichi FURUYA  Yoichi HANEDA  Akitoshi KATAOKA 

Publication
A - Abstracts of IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences (Japanese Edition)  Vol.J93-A  No.6  pp.387-396
Publication Date: 2010/06/01
Online ISSN: 1881-0195
Print ISSN: 0913-5707
Type of Manuscript: PAPER
Category: 
Keyword: 
dereverberation,  noise reduction,  blind deconvolution,  spectral subtraction,  speech enhancement, 

Full Text(in Japanese): PDF(986KB)

Summary: 
A speech enhancement method is presented for noisy reverberant conditions, such as hands-free teleconferencing and videoconferencing. This method introduces the constraint of minimizing the noise power in the inverse filter computation of dereverberation. The soft-constrained inverse filter reduces reverberation and noise and improves speech quality. A tradeoff is shown to exit between reducing the reverberation and reducing the noise; this tradeoff can be controlled by the constraint. The performance of our method in noisy reverberant conditions was objectively and subjectively evaluated in experiments using measured room impulse responses. The results indicate that this method provides better speech quality than conventional methods.