Multi-Scale Spectrogram Modelling for Neural Text-to-Speech

TL;DR

This paper introduces a multi-scale spectrogram modeling approach for neural text-to-speech synthesis, improving prosody by predicting speech features at multiple linguistic scales, leading to better speech quality.

Contribution

The paper presents a novel multi-scale spectrogram prediction mechanism with two versions, Word-level MSS and Sentence-level MSS, leveraging linguistic units for enhanced prosody modeling.

Findings

Word-level MSS outperforms baseline in subjective evaluations.

Multi-scale approach captures both coarse and fine prosody.

Linguistically motivated scales improve speech synthesis quality.

Abstract

We propose a novel Multi-Scale Spectrogram (MSS) modelling approach to synthesise speech with an improved coarse and fine-grained prosody. We present a generic multi-scale spectrogram prediction mechanism where the system first predicts coarser scale mel-spectrograms that capture the suprasegmental information in speech, and later uses these coarser scale mel-spectrograms to predict finer scale mel-spectrograms capturing fine-grained prosody. We present details for two specific versions of MSS called Word-level MSS and Sentence-level MSS where the scales in our system are motivated by the linguistic units. The Word-level MSS models word, phoneme, and frame-level spectrograms while Sentence-level MSS models sentence-level spectrogram in addition. Subjective evaluations show that Word-level MSS performs statistically significantly better compared to the baseline on two voices.