Multi-View Attention Transfer for Efficient Speech Enhancement

TL;DR

This paper introduces multi-view attention transfer (MV-AT), a feature-based knowledge distillation method for efficient, low-complexity speech enhancement models that maintain high performance in the time domain.

Contribution

It proposes a novel MV-AT method that transfers multi-view knowledge without extra parameters, improving lightweight models for speech enhancement.

Findings

Significant reduction in parameters and FLOPs with maintained performance.

Consistent performance improvements across various model sizes.

Effective knowledge transfer demonstrated on Valentini and DNS datasets.

Abstract

Recent deep learning models have achieved high performance in speech enhancement; however, it is still challenging to obtain a fast and low-complexity model without significant performance degradation. Previous knowledge distillation studies on speech enhancement could not solve this problem because their output distillation methods do not fit the speech enhancement task in some aspects. In this study, we propose multi-view attention transfer (MV-AT), a feature-based distillation, to obtain efficient speech enhancement models in the time domain. Based on the multi-view features extraction model, MV-AT transfers multi-view knowledge of the teacher network to the student network without additional parameters. The experimental results show that the proposed method consistently improved the performance of student models of various sizes on the Valentini and deep noise suppression (DNS) datasets. MANNER-S-8.1GF with our proposed method, a lightweight model for efficient deployment, achieved 15.4x and 4.71x fewer parameters and floating-point operations (FLOPs), respectively, compared to the baseline model with similar performance.