The Effectiveness of Time Stretching for Enhancing Dysarthric Speech for Improved Dysarthric Speech Recognition

TL;DR

This study evaluates various methods, including GAN-based voice conversion and simple signal processing, for enhancing dysarthric speech to improve recognition accuracy, finding that straightforward techniques can be as effective as complex models.

Contribution

The paper introduces a comparative analysis of GAN-based and signal processing methods, proposing a combined approach that enhances dysarthric speech recognition.

Findings

Simple signal processing methods achieve recognition results comparable to GAN-based methods.

Combining MaskCycleGAN-VC with time stretching improves recognition for some dysarthric speakers.

Straightforward techniques can be effective alternatives to complex GAN models.

Abstract

In this paper, we investigate several existing and a new state-of-the-art generative adversarial network-based (GAN) voice conversion method for enhancing dysarthric speech for improved dysarthric speech recognition. We compare key components of existing methods as part of a rigorous ablation study to find the most effective solution to improve dysarthric speech recognition. We find that straightforward signal processing methods such as stationary noise removal and vocoder-based time stretching lead to dysarthric speech recognition results comparable to those obtained when using state-of-the-art GAN-based voice conversion methods as measured using a phoneme recognition task. Additionally, our proposed solution of a combination of MaskCycleGAN-VC and time stretched enhancement is able to improve the phoneme recognition results for certain dysarthric speakers compared to our time stretched baseline.