English

Dynamically Slimmable Speech Enhancement Network with Metric-Guided Training

Audio and Speech Processing 2026-03-16 v3

Abstract

To further reduce the complexity of lightweight speech enhancement models, we introduce a gating-based Dynamically Slimmable Network (DSN). The DSN comprises static and dynamic components. For architecture-independent applicability, we introduce distinct dynamic structures targeting the commonly used components, namely, grouped recurrent neural network units, multi-head attention, convolutional, and fully connected layers. A policy module adaptively governs the use of dynamic parts at a frame-wise resolution according to the input signal quality, controlling computational load. We further propose Metric-Guided Training (MGT) to explicitly guide the policy module in assessing input speech quality. Experimental results demonstrate that the DSN achieves comparable enhancement performance in instrumental metrics to the state-of-the-art lightweight baseline, while using only 73% of its computational load on average. Evaluations of dynamic component usage ratios indicate that the MGT-DSN can appropriately allocate network resources according to the severity of input signal distortion.

Keywords

Cite

@article{arxiv.2510.11395,
  title  = {Dynamically Slimmable Speech Enhancement Network with Metric-Guided Training},
  author = {Haixin Zhao and Kaixuan Yang and Nilesh Madhu},
  journal= {arXiv preprint arXiv:2510.11395},
  year   = {2026}
}

Comments

Accepted by ICASSP2026

R2 v1 2026-07-01T06:34:00.318Z