PeriodGrad: Towards Pitch-Controllable Neural Vocoder Based on a Diffusion Probabilistic Model
This addresses the demand for pitch-controllable neural vocoders in practical applications such as singing voice synthesis, representing an incremental improvement over existing methods.
The paper tackles the problem of generating high-fidelity speech waveforms with flexible pitch control in neural vocoders, particularly for applications like singing voice synthesis, by incorporating explicit periodic signals into a diffusion probabilistic model. The result is improved sound quality and better pitch control compared to conventional DDPM-based neural vocoders.
This paper presents a neural vocoder based on a denoising diffusion probabilistic model (DDPM) incorporating explicit periodic signals as auxiliary conditioning signals. Recently, DDPM-based neural vocoders have gained prominence as non-autoregressive models that can generate high-quality waveforms. The neural vocoders based on DDPM have the advantage of training with a simple time-domain loss. In practical applications, such as singing voice synthesis, there is a demand for neural vocoders to generate high-fidelity speech waveforms with flexible pitch control. However, conventional DDPM-based neural vocoders struggle to generate speech waveforms under such conditions. Our proposed model aims to accurately capture the periodic structure of speech waveforms by incorporating explicit periodic signals. Experimental results show that our model improves sound quality and provides better pitch control than conventional DDPM-based neural vocoders.