一种高效的语音关键词检索系统.pdfVIP

第 27 卷第 2 期 通 信 学 报 Vol.27 No.2 2006 年 2 月 Journal on Communications February 2006 一种高效的语音关键词检索系统 罗骏, 欧智坚 (清华大学 电子工程系，北京 100084) 摘 要：针对音频检索任务中的关键词检索提出一种新的基于拼音图的两阶段检索系统，可以高效地从大量语音 数据中检索出感兴趣的文本信息，从而达到为国家安全服务的目的。该系统分为预处理阶段和检索阶段。预处理 阶段将语音数据识别成具有高覆盖率的拼音图，在这一过程中通过若干次的无监督最大似然线性回归自适应算法 渐次提高拼音图的质量。检索阶段响应用户的频繁查询，只需在拼音图中查找出与关键词拼音匹配的拼音串，并 采用基于 N 元拼音文法的前后向算法计算置信度以实现对检索结果的筛选。实验表明：系统具有较高的召回率和 正确率，且检索阶段仅需 0.01 倍实时，可以满足快速检索的需要。 关键词：信息检索；关键词检索；拼音图；置信度 中图分类号：TP391 文献标识码：B 文章编号：1000-436X(2006)02-0113-06 Efficient keyword spotting system for information retrieval LUO Jun, OU Zhi-jian (Department of Electronic Engineering, Tsinghua University, Beijing 100084, China) Abstract: A new two-stage keyword spotting system was proposed based on syllable graph for audio information retrieval task, which could efficiently spot the interesting words in mass speech data, thus serve for the national security. It comprised two stages – preprocessing and searching. In the preprocessing stage, the audio data was recognized into syllable graph which included high accuracy syllable candidates, and unsupervised MLLR (maximum likelihood linear regression) adaptation was carried out iteratively to further improve the accuracy of the syllable graph. In the searching stage, to answer the frequent queries from users, searching for matched keywords was only scanned in the graph for likely syllable strings. A forward-backward algorithm based on syl