Using StarGANv2 Voice Conversion to Enhance the Quality of Dysarthric Speech

In this paper, we propose to use StarGAN, a pow-erful Generative Adversarial Network (GAN), to improve the quality of dysarthric speech. Through extensive experiments, we demonstrate the effectiveness of StarGANv2-VC in converting dysarthric speech and significantly improving its intelligibility and naturalness. In addition, this research contributes to the field by conducting a comparative study between StarGANv2-VC and MaskCycleGAN-VC, another well-established GAN architecture, recently used in dysarthric speech conversion tasks. The results show that StarGANv2-VC performs the best, making it a promising solution for improving the speech quality of people suffering from dysarthria.