Learning Neural Vocoder from Range-Null Space Decomposition

TL;DR

This paper introduces a novel neural vocoder leveraging range-null space decomposition in the time-frequency domain, improving spectral detail generation and achieving state-of-the-art results with lightweight models.

Contribution

It proposes a new dual-path neural vocoder framework based on classical RND theory, combining linear domain shift and learnable spectral detail modeling.

Findings

Achieves state-of-the-art performance on LJSpeech and LibriTTS benchmarks.

Uses lightweight network parameters while maintaining high quality.

Demonstrates effective spectral detail reconstruction via range-null space decomposition.

Abstract

Despite the rapid development of neural vocoders in recent years, they usually suffer from some intrinsic challenges like opaque modeling, and parameter-performance trade-off. In this study, we propose an innovative time-frequency (T-F) domain-based neural vocoder to resolve the above-mentioned challenges. To be specific, we bridge the connection between the classical signal range-null decomposition (RND) theory and vocoder task, and the reconstruction of target spectrogram can be decomposed into the superimposition between the range-space and null-space, where the former is enabled by a linear domain shift from the original mel-scale domain to the target linear-scale domain, and the latter is instantiated via a learnable network for further spectral detail generation. Accordingly, we propose a novel dual-path framework, where the spectrum is hierarchically encoded/decoded, and the cross- and narrow-band modules are elaborately devised for efficient sub-band and sequential modeling. Comprehensive experiments are conducted on the LJSpeech and LibriTTS benchmarks. Quantitative and qualitative results show that while enjoying lightweight network parameters, the proposed approach yields state-of-the-art performance among existing advanced methods. Our code and the pretrained model weights are available at https://github.com/Andong-Li-speech/RNDVoC.