Test-time adaptation for speech enhancement with an autoregressive speech prior

2026-09-03

AI Summary

研究提出一种单语句测试时自适应方法，利用自回归语音先验调整预训练的语音增强模型，以改善声学条件不匹配下的语音质量。

Abstract

Test-time adaptation (TTA) offers a promising direction for improving speech enhancement models under mismatched acoustic conditions, without requiring access to labeled target data. In this work, we propose a single-utterance TTA method that regularizes a pretrained speech enhancement model using an autoregressive prior trained on clean speech latent representations extracted from a neural audio codec. Adaptation is performed by minimizing the Kullback-Leibler divergence between the enhanced speech distribution and the clean speech prior. Experiments across multiple noisy speech datasets show consistent improvements in speech quality, particularly under training-testing noise mismatch conditions. Code and audio examples are available online.

Problem

• Test-time adaptation
• speech enhancement
• mismatched acoustic conditions

Innovation

• Test-time Adaptation
• Autoregressive Prior
• Speech Enhancement
• Kullback-Leibler Divergence