An integrated multi-channel approach for joint noise reduction and dereverberation. (1st January 2021)

• Record Type: Journal Article
• Authors: Song, Siyuan; Cheng, Longbiao; Luan, Shuming; Yao, Dingding; Li, Junfeng; Yan, Yonghong
• Is Part Of: Applied acoustics. Volume 171(2020)
• Publication Date: 2021-01-01
• Subjects: Speech enhancement -- Weighted prediction error -- Multi-channel Wiener filter -- Minimum variance distortionless response -- Dereverberation and noise reduction; Acoustical engineering -- Periodicals; Periodicals; 620.2
• DOI: 10.1016/j.apacoust.2020.107526
• Languages: English
• ISSNs: 0003-682X

Abstract

Highlights: Multi-channel approach for joint noise reduction and dereverberation. Solutions based on Multi-channel Wiener Filter and Weighted Prediction Error. Relative early transfer function using covariance whitening method. Feedback updates target signal variance using system output. Proposed approach outperforms state-of-art approaches in speech enhancement. Abstract: The performance of speech communication and automatic speech recognition systems greatly degrades in real conditions where reverberation and noise exist. In this paper, we propose an integrated multi-channel speech enhancement approach for joint noise reduction and dereverberation based on the multi-channel Wiener filter (MWF) and the weighted prediction error (WPE). Specifically, the input signal of MWF is derived from WPE that is considered as a preprocessor. MWF is implemented as a minimum variance distortionless response (MVDR) beamformer followed by a single-channel Wiener filter. The covariance matrix of the undesired signal in MWF is approximated using the prediction in WPE. In the implementation of MVDR beamformer, the relative early transfer functions (RETFs) are used rather than either the entire relative transfer functions (RTFs) or only the direct-path of desired speech signal, which are estimated by covariance whitening method. The target signal variance in WPE is further updated with the output signal. Experimental results show that this proposed approach outperforms the state-of-art approaches in reducing late reverberation and noise, with the reduction of word error rate of 38.50% in 3 m recorded data conditions.