Improve GAN-based Neural Vocoder using Pointwise Relativistic LeastSquare GAN

TL;DR

This paper introduces PRLSGAN, a novel relativistic LSGAN variant for neural vocoders, which improves waveform synthesis quality by considering score distribution and combining loss functions, applicable to existing GAN vocoders.

Contribution

The paper proposes PRLSGAN, a new relativistic LSGAN framework that enhances GAN-based neural vocoders' quality by incorporating pointwise relative discrepancy loss.

Findings

PRLSGAN improves waveform quality in Parallel WaveGAN and MelGAN.

The framework demonstrates consistent performance gains.

PRLSGAN shows strong generalization across different vocoders.

Abstract

GAN-based neural vocoders, such as Parallel WaveGAN and MelGAN have attracted great interest due to their lightweight and parallel structures, enabling them to generate high fidelity waveform in a real-time manner. In this paper, inspired by Relativistic GAN, we introduce a novel variant of the LSGAN framework under the context of waveform synthesis, named Pointwise Relativistic LSGAN (PRLSGAN). In this approach, we take the truism score distribution into consideration and combine the original MSE loss with the proposed pointwise relative discrepancy loss to increase the difficulty of the generator to fool the discriminator, leading to improved generation quality. Moreover, PRLSGAN is a general-purposed framework that can be combined with any GAN-based neural vocoder to enhance its generation quality. Experiments have shown a consistent performance boost based on Parallel WaveGAN and MelGAN, demonstrating the effectiveness and strong generalization ability of our proposed PRLSGAN neural vocoders.