SEMamba++: A General Speech Restoration Framework Leveraging Global, Local, and Periodic Spectral Patterns

TL;DR

SEMamba++ is a novel speech restoration framework that integrates spectral periodicity and multi-resolution frequency analysis to improve performance on complex speech distortions efficiently.

Contribution

It introduces Frequency GLP and a multi-resolution dual-processing block, incorporating speech-specific features as inductive biases for enhanced restoration.

Findings

Achieves state-of-the-art performance on speech restoration benchmarks.

Maintains computational efficiency compared to baseline models.

Effectively leverages spectral periodicity and multi-resolution analysis.

Abstract

General speech restoration demands techniques that can interpret complex speech structures under various distortions. While State-Space Models like SEMamba have advanced the state-of-the-art in speech denoising, they are not inherently optimized for critical speech characteristics, such as spectral periodicity or multi-resolution frequency analysis. In this work, we introduce an architecture tailored to incorporate speech-specific features as inductive biases. In particular, we propose Frequency GLP, a frequency feature extraction block that effectively and efficiently leverages the properties of frequency bins. Then, we design a multi-resolution parallel time-frequency dual-processing block to capture diverse spectral patterns, and a learnable mapping to further enhance model performance. With all our ideas combined, the proposed SEMamba++ achieves the best performance among multiple baseline models while remaining computationally efficient.