[1]张亚歌张太镒夏川.噪声评估在端点检测中的应用[J].计算机技术与发展,2010,(09):177-180.
　ZHANG Ya-ge,ZHANG Tai-yi,XIA Chuan.Application of Noise Evaluation in Endpoint Detection[J].,2010,(09):177-180.
点击复制

噪声评估在端点检测中的应用()

分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

Author(s):: ZHANG Ya-ge; ZHANG Tai-yi; XIA Chuan; School of Electronic and Information Engineering,Xi＇an Jiaotong University

摘要:: 端点检测是语音识别中非常重要的部分,其准确性直接影响语音识别系统的识别率。传统端点检测方法预设经验门限对语音的短时特征进行判决,因为预设门限难以适应不同环境,其准确度和噪声鲁棒性较差。为了改善上述缺点,提出噪声评估的概念,对环境噪声的短时能量与短时过零率等短时特征进行分析,得到了更能表征环境噪声的门限。噪声评估结合传统的双门限法用于端点检测过程,解决了经验门限对不同环境适应性不强的问题。实验表明,噪声评估增加了端点检测的准确度和噪声鲁棒性

Abstract:: Endpoint detection is an important part in speech recognition and its accuracy affects the recognition rate of the whole recognition system directly. The traditional endpoint detection method uses empirical thresholds which are preset to judge the short-time characteristics of speech,preset thresholds are not accurate and noisy robust because they couldn＇t adapt to the environment.Therefore the notion of noise evaluation is proposed in this paper,evaluate the short-time characteristics of environmental noise such as short-time energy and short-time zero-cross-rate to get the thresholds that can better describe the environmental noise.Combining the noise evaluation with traditional two-threshold method in endpoint detection,the inelasticity of empirical thresholds is improved.Experimental results show that noise evaluation increases the accuracy and noisy robust of endpoint detection

[1]董胡. 低信噪比环境下改进的语音端点检测算法[J].计算机技术与发展,2016,26(03):71.
　DONG Hu. Improved Speech Endpoint Detection under Low SNR Environment[J].,2016,26(09):71.
[2]董胡. 基于先验信噪比和能零熵的语音端点检测算法[J].计算机技术与发展,2017,27(07):72.
　DONG Hu. Endpoint Detection Algorithm with Priori SNR and Energy-zero-spectral Entropy[J].,2017,27(09):72.
[3]于阿强,汪方军,孙存,等. 基于Android平台的吉他智能校准系统设计与实现[J].计算机技术与发展,2017,27(07):140.
　YU A-qiang,WANG Fang-jun,SUN Cun,et al. Design and Implementation of a Guitar Intelligent Calibration System with Android Platform[J].,2017,27(09):140.

备注/Memo:: 张亚歌（1984-），男，河南郑州人，硕士研究生，研究方向为语音信号处理、语音识别；张太镒，博士生导师，研究方向为新一代移动通信技术、软件无线电，数字音视频技术

更新日期/Last Update: 1900-01-01