MetricGAN+: An Improved Version of MetricGAN for Speech Enhancement

TL;DR

This paper introduces MetricGAN+, an enhanced speech enhancement method that optimizes perceptual quality metrics directly, leading to improved speech quality and state-of-the-art PESQ scores on the VoiceBank-DEMAND dataset.

Contribution

MetricGAN+ incorporates three domain-knowledge-based training techniques to improve upon the original MetricGAN for better speech enhancement performance.

Findings

PESQ score increased by 0.3 over previous MetricGAN

Achieved state-of-the-art PESQ score of 3.15

Demonstrated effectiveness on VoiceBank-DEMAND dataset

Abstract

The discrepancy between the cost function used for training a speech enhancement model and human auditory perception usually makes the quality of enhanced speech unsatisfactory. Objective evaluation metrics which consider human perception can hence serve as a bridge to reduce the gap. Our previously proposed MetricGAN was designed to optimize objective metrics by connecting the metric with a discriminator. Because only the scores of the target evaluation functions are needed during training, the metrics can even be non-differentiable. In this study, we propose a MetricGAN+ in which three training techniques incorporating domain-knowledge of speech processing are proposed. With these techniques, experimental results on the VoiceBank-DEMAND dataset show that MetricGAN+ can increase PESQ score by 0.3 compared to the previous MetricGAN and achieve state-of-the-art results (PESQ score = 3.15).