Combination of MVDR beamforming and single-channel spectral processing for enhancing noisy and reverberant speech

Abstract

This paper presents a system aiming at joint dereverberation and noise reduction by applying a combination of a beamformer with a single-channel spectral enhancement scheme. First, a minimum variance distortionless response beamformer with an online estimated noise coherence matrix is used to suppress noise and reverberation. The output of this beamformer is then processed by a single-channel spectral enhancement scheme, based on statistical room acoustics, minimum statistics, and temporal cepstrum smoothing, to suppress residual noise and reverberation. The evaluation is conducted using the REVERB challenge corpus, designed to evaluate speech enhancement algorithms in the presence of both reverberation and noise. The proposed system is evaluated using instrumental speech quality measures, the performance of an automatic speech recognition system, and a subjective evaluation of the speech quality based on a MUSHRA test. The performance achieved by beamforming, single-channel spectral enhancement, and their combination are compared, and experimental results show that the proposed system is effective in suppressing both reverberation and noise while improving the speech quality. The achieved improvements are particularly significant in conditions with high reverberation times

Joint Late Reverberation and Noise Power Spectral Density Estimation in a Spatially Homogeneous Noise Field

“…Dereverberation and noise reduction techniques aim at designing the filter w(l) such that the output signal Z(l) is as close as possible to the target signal S(l). A widely used dereverberation and noise reduction technique is the MWF, which aims at minimizing the meansquare error between Z(l) and S(l) [4][5][6]. The MWF is typically implemented as a minimum variance distortionless response (MVDR) beamformer w MVDR (l) followed by a single-channel Wiener postfilter G(l) [10, 12-18], i.e.,…”

“…A commonly used dereverberation and noise reduction technique is the multi-channel Wiener filter (MWF), which aims at minimizing the mean-square error between the output signal and the target signal [4][5][6]. The implementation of the MWF requires (among other parameters) an estimate of the late reverberation and noise power spectral densities (PSDs).…”

Robust Speech Dereverberation With a Neural Network-Based Post-Filter That Exploits Multi-Conditional Training of Binaural Cues

“…In the context of single-channel dereverberation, this gain function is typically based on an estimate of the late reverberant power spectral density (PSD) [10], [11]. More sophisticated approaches aim at reducing both interfering noise and reverberation by either jointly estimating the noise and the late reverberant PSD [12] or by combining an autoregressive moving-average (ARMA) model of the late reverberant PSD with a hidden Markov model (HMM) of clean speech in a Bayesian filtering framework [13].…”

Under-modelled blind system identification for time delay estimation in reverberant environments

“…Reliable DOA estimation is important on applications such as speaker tracking [1,2], beamforming-based multichannel speech enhancement [3,4,5] and dereverberation [6,7].…”