From KAN to GR-KAN: Advancing Speech Enhancement with KAN-Based Methodology

TL;DR

This paper introduces GR-KAN, an improved KAN variant for speech enhancement that enhances model expressiveness and scalability, leading to better performance with fewer parameters in both time and frequency domains.

Contribution

It adapts GR-KAN to DNN-based speech enhancement, demonstrating significant parameter reduction and performance improvement over existing methods.

Findings

GR-KAN reduces parameters by up to 4x.

GR-KAN improves PESQ by up to 0.1.

First successful application of KAN-based methods in speech enhancement.

Abstract

Deep neural network (DNN)-based speech enhancement (SE) usually uses conventional activation functions, which lack the expressiveness to capture complex multiscale structures needed for high-fidelity SE. Group-Rational KAN (GR-KAN), a variant of Kolmogorov-Arnold Networks (KAN), retains KAN's expressiveness while improving scalability on complex tasks. We adapt GR-KAN to existing DNN-based SE by replacing dense layers with GR-KAN layers in the time-frequency (T-F) domain MP-SENet and adapting GR-KAN's activations into the 1D CNN layers in the time-domain Demucs. Results on Voicebank-DEMAND show that GR-KAN requires up to 4x fewer parameters while improving PESQ by up to 0.1. In contrast, KAN, facing scalability issues, outperforms MLP on a small-scale signal modeling task but fails to improve MP-SENet. We demonstrate the first successful use of KAN-based methods for consistent improvement in both time- and SoTA TF-domain SE, establishing GR-KAN as a promising alternative for SE.