arXiv cs.AI· Zhiwei Lin, Kaiqi Fu, Rime Wen, Zehan Liu, Shawn Qin, Roy Gan, Hao Wang·· 9 天前AI 评分24
X2Streaming-ASR：不确定就等待，准备好就输出，实现低延迟流式 ASR
X2Streaming-ASR: wait when uncertain, emit when ready for streaming ASR
AI 导读
X2Streaming-ASR 将流式语音识别拆解为"何时提交"与"提交什么"，通过三阶段训练（先建立流式识别能力，再用自动探测轨迹热启动提交策略，最后用字符级、分段分配的分组相对奖励优化准确率与延迟）。在十个中英文测试集上，其平均提交延迟相对强制对齐端点最低，中文为 32–109ms、英文为 12–85ms，识别准确率与现有系统相当。
正文 · 原文
Abstract:Streaming automatic speech recognition (ASR) for real-time voice agents and full-duplex dialogue must provide accurate partial transcripts with low commit latency. Existing systems commonly use a fixed chunk size, look-ahead, or target delay, or encourage emissions near estimated acoustic boundaries. These approaches do not directly optimize how much additional context to use at each output position under a single-pass, hard-commit constraint. We propose X2Streaming-ASR, which decomposes streaming recognition into when to commit and what to commit. Its three-stage training procedure first establishes streaming recognition ability, then warm-starts the commit policy with automatically probed trajectories, and finally refines the policy using character-level, segment-assigned group-relative rewards for recognition accuracy and latency. Across ten Chinese and English test sets, X2Streaming-ASR attains the lowest mean commit latency relative to forced-aligned endpoints, 32--109ms on Chinese characters and 12--85ms on English words, while recognition accuracy remains comparable to existing systems.
| Subjects: | Sound (cs.SD); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.08672 [cs.SD] |
| (or arXiv:2609.08672v3 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2609.08672 arXiv-issued DOI via DataCite |
Submission history
From: Zhiwei Lin [view email]
[v1]
Tue, 8 Sep 2026 12:39:06 UTC (149 KB)
[v2]
Tue, 15 Sep 2026 09:12:31 UTC (152 KB)
[v3]
Thu, 8 Oct 2026 06:19:33 UTC (157 KB)
来源：arXiv cs.AI · arxiv.org