Real-time multichannel deep speech enhancement in hearing aids: Comparing monaural and binaural processing in complex acoustic scenarios

TL;DR

This paper compares monaural and binaural deep learning-based speech enhancement algorithms for hearing aids, demonstrating binaural processing's advantages in complex acoustic environments with spatial interference.

Contribution

It introduces and evaluates low-complexity deep speech enhancement algorithms for hearing aids, contrasting monaural and binaural approaches in real-world scenarios.

Findings

Binaural deep learning outperforms monaural in spatial interference conditions.

All algorithms perform similarly in diffuse noise environments.

Binaural approach shows significant improvements at low SNRs.

Abstract

Deep learning has the potential to enhance speech signals and increase their intelligibility for users of hearing aids. Deep models suited for real-world application should feature a low computational complexity and low processing delay of only a few milliseconds. In this paper, we explore deep speech enhancement that matches these requirements and contrast monaural and binaural processing algorithms in two complex acoustic scenes. Both algorithms are evaluated with objective metrics and in experiments with hearing-impaired listeners performing a speech-in-noise test. Results are compared to two traditional enhancement strategies, i.e., adaptive differential microphone processing and binaural beamforming. While in diffuse noise, all algorithms perform similarly, the binaural deep learning approach performs best in the presence of spatial interferers. Through a post-analysis, this can be attributed to improvements at low SNRs and to precise spatial filtering.