提出全局-局部融合的专家路由机制，提升重叠语音识别准确率。
GLAD: Global-Local Aware Dynamic Mixture-of-Experts for Multi-Talker ASR
- 动态融合说话人全局上下文与局部声学特征，指导专家选择。
- 在LibriSpeechMix和CH109数据集上显著优于现有SOT方法。
- 特别适合高重叠复杂场景，首次将全局-局部MoE用于多说话人识别。
端到端多说话人自动语音识别（MTASR）在处理重叠语音时面临巨大挑战，关键瓶颈在于说话人特有的声学特征在深层网络中常被稀释。为此，本文提出全局-局部感知动态专家混合（GLAD）架构。GLAD引入新型路由机制，动态融合说话人感知的全局上下文与精细局部声学细节，自适应地引导专家选择。在LibriSpeechMix和CH109数据集上的实验表明，GLAD显著优于现有的基于序列输出训练（SOT）的MTASR方法，在高重叠复杂场景下表现出卓越鲁棒性。据我们所知，这是首个将全局-局部融合的MoE策略应用于MTASR的工作。
原文摘要 · Abstract (English)
End-to-end multi-talker automatic speech recognition (MTASR) faces significant challenges in accurately transcribing overlapping speech. A critical bottleneck is that speaker-specific acoustic characteristics, which are essential for distinguishing overlapping speech, are often diluted in deep network layers. To address this, we propose the Global-Local Aware Dynamic Mixture-of-Experts (GLAD) architecture. GLAD introduces a novel routing mechanism that dynamically fuses speaker-aware global context with fine-grained local acoustic details to adaptively guide expert selection. Experiments on the LibriSpeechMix and CH109 datasets demonstrate that GLAD significantly outperforms existing Serialized Output Training (SOT)-based MTASR approaches, exhibiting exceptional robustness in challenging, high-overlap scenarios. To the best of our knowledge, this is the first work to apply a global-local fusion MoE strategy to MTASR.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。