Generative Adversarial Networks for Unpaired Voice Transformation on Impaired Speech

TL;DR

This paper introduces a novel end-to-end GAN-based unsupervised voice conversion method that transforms impaired speech into normal speech, enhancing intelligibility without requiring parallel data, and outperforms existing CycleGAN techniques.

Contribution

It presents the first end-to-end GAN model for unsupervised voice conversion specifically targeting impaired speech, preserving linguistic content and speaker identity.

Findings

Outperforms CycleGAN in transforming impaired speech

Preserves linguistic content and speaker characteristics

Effective for improving speech intelligibility in surgical patients

Abstract

This paper focuses on using voice conversion (VC) to improve the speech intelligibility of surgical patients who have had parts of their articulators removed. Due to the difficulty of data collection, VC without parallel data is highly desired. Although techniques for unparallel VC, for example, CycleGAN, have been developed, they usually focus on transforming the speaker identity, and directly transforming the speech of one speaker to that of another speaker and as such do not address the task here. In this paper, we propose a new approach for unparallel VC. The proposed approach transforms impaired speech to normal speech while preserving the linguistic content and speaker characteristics. To our knowledge, this is the first end-to-end GAN-based unsupervised VC model applied to impaired speech. The experimental results show that the proposed approach outperforms CycleGAN.