Microphone Array Signal Processing and Deep Learning for Speech Enhancement

TL;DR

This paper compares model-based, data-driven, and hybrid methods for multi-channel speech enhancement, highlighting how combining these approaches can improve noise reduction, source separation, and dereverberation.

Contribution

It introduces a hybrid approach that leverages both model-based and deep learning techniques for improved parameter estimation in speech enhancement.

Findings

Hybrid methods outperform purely model-based or data-driven approaches.

Deep learning enhances the estimation of spatial filtering parameters.

Hybrid approaches effectively address noise and reverberation challenges.

Abstract

Multi-channel acoustic signal processing is a well-established and powerful tool to exploit the spatial diversity between a target signal and non-target or noise sources for signal enhancement. However, the textbook solutions for optimal data-dependent spatial filtering rest on the knowledge of second-order statistical moments of the signals, which have traditionally been difficult to acquire. In this contribution, we compare model-based, purely data-driven, and hybrid approaches to parameter estimation and filtering, where the latter tries to combine the benefits of model-based signal processing and data-driven deep learning to overcome their individual deficiencies. We illustrate the underlying design principles with examples from noise reduction, source separation, and dereverberation.