Statistical Parametric Speech Synthesis Using Generative Adversarial Networks Under A Multi-task Learning Framework

TL;DR

This paper introduces a multi-task learning framework combining traditional acoustic loss and GAN discriminative loss to enhance the naturalness of synthesized speech in statistical parametric speech synthesis.

Contribution

It proposes a novel GAN-based architecture integrated with MSE loss under a multi-task learning framework for improved speech synthesis quality.

Findings

Generated speech is more natural according to listening tests.

The multi-task framework stabilizes GAN training.

Synthesized speech better matches natural speech distribution.

Abstract

In this paper, we aim at improving the performance of synthesized speech in statistical parametric speech synthesis (SPSS) based on a generative adversarial network (GAN). In particular, we propose a novel architecture combining the traditional acoustic loss function and the GAN's discriminative loss under a multi-task learning (MTL) framework. The mean squared error (MSE) is usually used to estimate the parameters of deep neural networks, which only considers the numerical difference between the raw audio and the synthesized one. To mitigate this problem, we introduce the GAN as a second task to determine if the input is a natural speech with specific conditions. In this MTL framework, the MSE optimization improves the stability of GAN, and at the same time GAN produces samples with a distribution closer to natural speech. Listening tests show that the multi-task architecture can generate more natural speech that satisfies human perception than the conventional methods.