Perceptually Better, Semantically Worse: Measuring Speech Enhancement Impact on LLM-Based Voice Systems

2026-08-31

AI Summary

研究探讨了语音增强对基于大语言模型的语音系统下游任务性能的影响，通过引入输出分歧率(ODR)测量方法，并在多种条件下进行测试。

Abstract

Speech enhancement (SE) is commonly applied as a preprocessing step in spoken AI pipelines under the assumption that better audio quality improves downstream task performance. Whether SE-induced distortions propagate to downstream LLM task performance remains an open question. We introduce Output Divergence Rate (ODR), which measures how often SE changes an LLM's intent classification relative to clean speech, and benchmark five conditions on 2,974 SLURP clips using Whisper large-v3 and wav2vec2-large cascades. Every condition produces ODR significantly above zero ($p < 0.001$, binomial test). MetricGAN{+} more than doubles ODR versus unenhanced noisy speech (0.318 vs. 0.135) despite improving PESQ, and unmitigated echo reaches an ODR of 0.836 through speaker substitution, a failure WER cannot capture. Audio quality metrics range from near-zero to moderate correlation with ODR (SQUIM-MOS $ρ=-0.068$, PESQ $ρ=-0.467$). The MetricGAN{+} and echo results replicate across ASR architectures, indicating that standard audio quality metrics are insufficient for LLM pipeline quality.

Problem

• Speech Enhancement
• Large Language Models
• Output Divergence Rate
• Audio Quality Metrics

Innovation

• Output Divergence Rate
• speech enhancement
• large language model
• intent classification
• audio quality metrics