SALTTS: Leveraging Self-Supervised Speech Representations for improved Text-to-Speech Synthesis

TL;DR

This paper introduces SALTTS, a novel TTS method that incorporates self-supervised speech representations to enhance speech quality, outperforming baseline models through both objective and subjective evaluations.

Contribution

It proposes leveraging SSL speech representations within FastSpeech2, introducing parallel and cascade architectures for improved synthesis quality.

Findings

Enhanced speech naturalness and clarity in synthesized output.

Objective and subjective metrics show significant improvements over baseline.

SSL features contribute to richer and more accurate speech characteristics.

Abstract

While FastSpeech2 aims to integrate aspects of speech such as pitch, energy, and duration as conditional inputs, it still leaves scope for richer representations. As a part of this work, we leverage representations from various Self-Supervised Learning (SSL) models to enhance the quality of the synthesized speech. In particular, we pass the FastSpeech2 encoder's length-regulated outputs through a series of encoder layers with the objective of reconstructing the SSL representations. In the SALTTS-parallel implementation, the representations from this second encoder are used for an auxiliary reconstruction loss with the SSL features. The SALTTS-cascade implementation, however, passes these representations through the decoder in addition to having the reconstruction loss. The richness of speech characteristics from the SSL features reflects in the output speech quality, with the objective and subjective evaluation measures of the proposed approach outperforming the baseline FastSpeech2.