arXiv cs.CL· Husein Zolkepli·· 13 小时前AI 评分52
论文提出因果语音隔离器，开源模型在 AMI 通话上词错率 16.13%
The Voice on the Line
AI 导读
论文训练了 140M、243M 和 374M 参数的因果语音隔离器，延迟 20 ms，并从通话本身注册来电者声音。作者称发布版在 60 秒 AMI 窗口上词错率为 16.13%，低于 VoxLock 的 23.66%，在 Krisp 基准上与对手持平（5.74% 对 5.63%）。
正文 · 原文
Abstract:A transcriber listening through one microphone also hears everyone near it. We train causal voice isolators: masking transformers of 140M, 243M and 374M parameters, 20 ms of latency, enrolling the caller from the call itself. Short evaluation windows hide two failures that real calls expose: forgetting the caller when another voice talks alone for a while, and damaging a caller who is alone. Long calls with turns, memory beyond the attention window, real headset crosstalk and drifting caller levels address both, but each trades one against the other. WavLM-Large's own weights, made causal and trained with the model, move the trade: M with them (555M parameters) lowers the error on our development sets from 11.44% to 10.55%; a student trained from scratch to imitate WavLM does not, and w2v-BERT 2.0 and Whisper's encoder, pre-trained on far more audio, help less or not at all. A lone caller loses most at the start of a recording: no training call opened with someone already talking. Training on windows cut from longer calls lowers the error on clean development calls from 3.92% to 2.96%; this is our release. Live, the release makes 16.13% word errors on 60 s AMI windows, against 23.66% for VoxLock, the strongest open isolator, and 10.19% on its own AMI windows, against 11.71%; it is level on Krisp's benchmark (5.74% against 5.63%), and a caller who is alone still loses more words than with VoxLock (5.58% against 2.92% on clean test calls). Distilled into 22.1M parameters without WavLM, it runs live on one CPU thread inside a LiveKit agent, at 33% of a core per call in 20 ms frames, with 10.66% error on our development sets (the release 10.49%). We open-source the release and this CPU model, with a LiveKit plugin, at this https URL.
| Subjects: | Sound (cs.SD); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.35272 [cs.SD] |
| (or arXiv:2610.35272v1 [cs.SD] for this version) | |
| https://doi.org/10.48550/arXiv.2610.35272 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Husein Zolkepli [view email]
[v1]
Mon, 26 Oct 2026 11:46:17 UTC (4,201 KB)
来源：arXiv cs.CL · arxiv.org