Integrated Sidelobe Cancellation and Linear Prediction Kalman Filter for Joint Multi-Microphone Speech Dereverberation, Interfering Speech Cancellation, and Noise Reduction

Abstract

In multi-microphone speech enhancement, reverberation as well as additive noise and/or interfering speech are commonly suppressed by deconvolution and spatial filtering, e.g., using multi-channel linear prediction (MCLP) on the one hand and beamforming, e.g., a generalized sidelobe canceler (GSC), on the other hand. In this paper, we consider several reverberant speech components, whereof some are to be dereverberated and others to be canceled, as well as a diffuse (e.g., babble) noise component to be suppressed. In order to perform both deconvolution and spatial filtering, we integrate MCLP and the GSC into a novel architecture referred to as integrated sidelobe cancellation and linear prediction (ISCLP), where the sidelobecancellation (SC) filter and the linear prediction (LP) filter operate in parallel, but on different microphone signal frames. Within ISCLP, we estimate both filters jointly by means of a single Kalman filter. We further propose a spectral Wiener gain post-processor, which is shown to relate to the Kalman filter's posterior state estimate. The presented ISCLP Kalman filter is benchmarked against two state-of-the-art approaches, namely first a pair of alternating Kalman filters respectively performing dereverberation and noise reduction, and second an MCLP+GSC Kalman filter cascade. While the ISCLP Kalman filter is roughly M 2 times less expensive than both reference algorithms, where M denotes the number of microphones, it is shown to perform similarly as compared to the former, and to outperform the latter.

Square Root-Based Multi-Source Early PSD Estimation and Recursive RETF Update in Reverberant Environments by Means of the Orthogonal Procrustes Problem

“…Sec. VI-A3, the penalty factor α in the conventional MP in (23) and the square-root MP in (34) is simulated in the range α ∈ [10 −3 , 10 5 ]. We perform at most i max = 20 iterations of the associated iterative algorithms in (24)- (25) and (35)- (36).…”

“…is not required in(34), as in the square-root signal model, we find that ϕ s = Diag[ϕ , and therefore the corresponding estimateφ s is guaranteed to be non-negative. Problems of the kind as in (34), i.e.…”

“…Regardless, we simulate frequencies up to f = 8 kHz, corresponding to a virtual sampling frequency of f s = 16 kHz. The soft-constraint penalty factor α in the conventional MP in(23) and the square-root MP in(34) is simulated in the range α ∈ [10 −3 , 10 5 ]. We perform at most i max = 20 iterations of the associated iterative algorithms in(24)-(25) and (35)-ε ϕs versus f for conventional MP [ ] and squareroot MP [ ] with α = 10 3 at (a) ε H = 0 dB, (b) ε H = −10 dB, and (c) ε H = −20 dB.…”

A CNN-based approach to identification of degradations in speech signals

“…A variety of effective signal enhancement techniques have been developed to enhance a degraded speech signal such as noise reduction [5,6], dereverberation [7,8], and restoration of some types of nonlinear distortion [9,10]. Most of these enhancement algorithms have been designed to deal with a specific type of degradation in a signal, although recent research in comprehensive speech enhancement, dealing with both additive noise and reverberation, is promising [11][12][13]. Nevertheless, to properly compensate for the effects of degradations, it is necessary to know or obtain information about the presence and the type of degradations in speech signals.…”

Generating coherence-constrained multisensor signals using balanced mixing and spectrally smooth filters

“…This also avoids the inclusion of unwanted acoustic sources in the recordings. Examples of multi-channel dereverberation approaches that utilize synthetic noise signals with specific spatial properties can be found, e.g., in Dietzen et al (2020) and Th€ une and Enzner (2017). In Habets (2019, 2020), synthetic wind noise exhibiting the Corcos model was used to evaluate the performance of two multi-channel wind noise reduction algorithms by adding the noise to the speech signals.…”