Adversarial Training of Denoising Diffusion Model Using Dual Discriminators for High-Fidelity Multi-Speaker TTS

TL;DR

This paper introduces a novel adversarial training approach for diffusion-based multi-speaker TTS using dual discriminators, significantly improving speech quality and synthesis speed over existing models.

Contribution

It proposes a diffusion speech synthesis model with two discriminators, enhancing the distribution learning of the reverse process and generated data, leading to superior performance.

Findings

Outperforms state-of-the-art models like FastSpeech2 and DiffGAN-TTS in multiple metrics.

Achieves higher speech quality and synthesis speed.

Demonstrates effectiveness through both objective and subjective evaluations.

Abstract

The diffusion model is capable of generating high-quality data through a probabilistic approach. However, it suffers from the drawback of slow generation speed due to the requirement of a large number of time steps. To address this limitation, recent models such as denoising diffusion implicit models (DDIM) focus on generating samples without directly modeling the probability distribution, while models like denoising diffusion generative adversarial networks (GAN) combine diffusion processes with GANs. In the field of speech synthesis, a recent diffusion speech synthesis model called DiffGAN-TTS, utilizing the structure of GANs, has been introduced and demonstrates superior performance in both speech quality and generation speed. In this paper, to further enhance the performance of DiffGAN-TTS, we propose a speech synthesis model with two discriminators: a diffusion discriminator for learning the distribution of the reverse process and a spectrogram discriminator for learning the distribution of the generated data. Objective metrics such as structural similarity index measure (SSIM), mel-cepstral distortion (MCD), F0 root mean squared error (F0 RMSE), short-time objective intelligibility (STOI), perceptual evaluation of speech quality (PESQ), as well as subjective metrics like mean opinion score (MOS), are used to evaluate the performance of the proposed model. The evaluation results show that the proposed model outperforms recent state-of-the-art models such as FastSpeech2 and DiffGAN-TTS in various metrics. Our implementation and audio samples are located on GitHub.