Robust Constrained MFMVDR Filters for Single-Channel Speech Enhancement Based on Spherical Uncertainty Set

Abstract

Aiming at exploiting speech correlation across consecutive time-frames in the short-time Fourier transform domain, the multi-frame minimum variance distortionless response (MFMVDR) filter for single-channel speech enhancement has been proposed. The MFMVDR filter requires an accurate estimate of the normalized speech correlation vector in order to avoid speech distortion and artifacts. In this paper we investigate the potential of using robust MVDR filtering techniques to estimate the normalized speech correlation vector as the vector maximizing the total signal output power within a spherical uncertainty set, which corresponds to imposing a quadratic inequality constraint. Whereas the singlyconstrained (SC) MFMVDR filter only considers the quadratic inequality constraint to estimate the (non-normalized) speech correlation vector, the doubly-constrained (DC) MFMVDR filter integrates a linear normalization constraint into the optimization problem to directly estimate the normalized speech correlation vector. To set the upper bound of the quadratic inequality constraint for each time-frequency point, we propose to use a trained non-linear mapping function that depends on the a-priori signal-to-noise ratio (SNR). Experimental results for different speech signals, noisetypes and SNRs show that the proposed constrained approaches yield a more accurate estimate of the normalized speech correlation vector than a state-of-the-art maximum-likelihood (ML) estimator. An instrumental and a perceptual evaluation show that both constrained MFMVDR filters lead to less speech and noise distortion but a lower noise reduction than the ML-MFMVDR filter, where the DC-MFMVDR filter is preferred in terms of overall quality compared to the SC-MFMVDR and ML-MFMVDR filters.

Distributed multichannel Wiener filtering for speech enhancement in acoustic sensor networks

“…Herein, we present detailed calculations of the expressions of the IPMWF in Equation ( 18), DPMWF in Equation ( 29), and SDPMWF in Equation (36) (temporarily denote 𝛼N in Equation (36) as 𝛽). In accordance with the Lemma 1 and the fact that R xx = 𝜎 2 S aa H and 𝜆 = 𝜎 2 S a H R −1 vv a, using Equation ( 18), we obtain…”

“…Since a single microphone is difficult to acquire spatial information, its speech enhancement performance is limited. 1,2 To break through this limitation, microphone array based speech enhancement techniques have emerged. [3][4][5][6][7] The microphone array consists of some microphones arranged in a geometric topology and its enhancement performance can be significantly improved by taking advantage of temporal spectral and spatial information about the speech and noise sources.…”

“…Speech enhancement plays an important role in improving the quality and intelligibility of speech under many application scenarios, such as teleconference systems, human‐machine communication, and speech recognition. Since a single microphone is difficult to acquire spatial information, its speech enhancement performance is limited 1,2 . To break through this limitation, microphone array based speech enhancement techniques have emerged 3‐7 .…”