| 一种基于滑动窗口的语音端点检测算法 |
| |
| 引用本文： | 余洪涌,赵庆卫,颜永红.一种基于滑动窗口的语音端点检测算法[J].微计算机应用,2006,27(6):641-645. |
| |
| 作者姓名： | 余洪涌 赵庆卫 颜永红 |
| |
| 作者单位： | 1. 中科信利通信技术有限公司,北京,100080
2. 中国科学院声学研究所中科信利语音实验室,北京,100080 |
| |
| 基金项目： | 国家高技术研究发展计划(863计划);中国科学院"百人计划";中国科学院知识创新工程项目 |
| |
| 摘 要： | 提出了一种基于滑动窗口的综合语音端点检测方法。从提高系统鲁棒性角度出发，利用各种综合的算法思想方法提高端点检测的准确性，鲁棒性和提高语音识别系统的总体识别率。利用频域能量、回声消除、DTMF滤除、利用信噪比确定多种阈值用于判定、利用语音渡形特点进行粗调和精调等。试验表明该方法在语音端点检测的错误率和端点误差方面都显著减少了，系统鲁棒性得到明显提高。
|
| 关 键 词： | 语音识别 语音端点检测 滑动窗口 回声消除 |
| 收稿时间： | 2004-10-12 |
| 修稿时间： | 2004-10-12 |
One Voice Activity Detection Algorithm Base on Slip Window |
| |
| YU Hongyong,ZHAO Qingwei,YAN Yongyong.One Voice Activity Detection Algorithm Base on Slip Window[J].Microcomputer Applications,2006,27(6):641-645. |
| |
| Authors: | YU Hongyong ZHAO Qingwei YAN Yongyong |
| |
| Affiliation: | 1 Beijing Kexin Technology Co. Ltd, Beijing, 100080, China ; 2 ZhongKeXinLi Speech Lab, Institute of Acoustics, Chinese Academy of Sciences, Beijing, 100080,China |
| |
| Abstract: | This paper presents a new compositive VAD algorithm base on one kind of slip window. We use many meth ods to improve speech detection robustness and the performance of system. These methods include use signal spectral energy, echo cancel, DTMF filter, two SNR , various thresholds, voice waveform and so on. Experimental results demonstrate a higher robustness and lower error detection rate in contract to the algorithm which do not use these methods. |
| |
| Keywords: | Speech recognize voice activity detection slip window echo cancel |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
