End-to-End Neural Formant Synthesis Using Low-Dimensional Acoustic Parameters

Sumiharu Kobayashi, Tetsuo Kosaka, Takashi Nose

Abstract: Neural vocoders can synthesize high-quality speech waveforms from acoustic features, but they cannot control by acoustic parameters, such as F0 and formant frequencies. Although analysis-synthesis based on signal processing can be controlled using acoustic parameters, its speech quality is inferior to that of neural vocoders. This paper proposes End-to-End Neural Formant Synthesis for generating high-quality speech waveforms with controllable acoustic parameters from low-dimensional representations. We compared three models with different structures, and investigated their synthesis quality and controllability. Experimental results showed that the proposed method performed as well as or better than conventional methods in terms of speech quality and controllability.