DocumentCode
1689338
Title
Dual-microphone noise reduction for mobile phone application
Author
Zhong-Hua Fu ; Fan Fan ; Jia-Dian Huang
Author_Institution
Sch. of Comput. Sci., Northwestern Polytech. Univ., Xi´an, China
fYear
2013
Firstpage
7239
Lastpage
7243
Abstract
This paper addresses the dual-microphone noise reduction problem in mobile phone application. We propose to use the inter-microphone Posteriori SNR Difference (PSNRD) for Speech Presence Probability (SPP) estimation, which is more robust than the Power Level Difference (PLD) that is often used previously. Additionally, we use the recently reported multichannel minimum variance distortionless response (MVDR) filter for noise reduction. We divide the noises into quasi-stationary and transient components, and propose an SPP-based noise correlation matrix estimator. Analysis and experiments on data recorded in real environments verify the robustness of the PSNRD. The SPP-based estimator is more appropriate for the MVDR filter in tracking transient noise, and the proposed MVDR filter can lead to high noise reduction and small speech distortion.
Keywords
filtering theory; matrix algebra; microphones; mobile handsets; probability; signal denoising; speech processing; MVDR filter; PSNRD; SPP estimation; SPP-based noise correlation matrix estimator; dual-microphone noise reduction; intermicrophone posteriori SNR difference; mobile phone application; multichannel minimum variance distortionless response filter; quasistationary components; small speech distortion; speech presence probability estimation; transient components; transient noise tracking; Correlation; Microphones; Mobile handsets; Noise reduction; Signal to noise ratio; Speech; Dual-microphone; MVDR filter; noise reduction; speech presence probability;
fLanguage
English
Publisher
ieee
Conference_Titel
Acoustics, Speech and Signal Processing (ICASSP), 2013 IEEE International Conference on
Conference_Location
Vancouver, BC
ISSN
1520-6149
Type
conf
DOI
10.1109/ICASSP.2013.6639068
Filename
6639068
Link To Document