基于LMS自适应的改进端点检测算法  

Improved Endpoint Detection Algorithm Based on LMS Adaptive

作　　者：刁鸿鹄 毛强 章小兵 DIAO Honghu;MAO Qiang;ZHANG Xiaobing(School of Electrical and Information Engineering,Anhui University of Technology,Maanshan 243032)

机构地区：[1]安徽工业大学电气与信息工程学院,安徽马鞍山243032

出　　处：《常州工学院学报》2021年第1期31-40,共10页Journal of Changzhou Institute of Technology

摘　　要：语音识别技术早已不再局限于实验室中,但部分关键技术仍需进一步优化与改进,以适应社会和用户不断增加的需求。从语音识别的技术理论入手,着重研究端点检测技术,拟在此基础上针对现有算法进行改进,提升其在低信噪比环境下的识别准确率。提出了一种在对语音进行LMS自适应降噪后,采用倒谱距离熵比的新型端点检测算法,并在MATLAB平台上进行相应测试。结果表明该算法相比于传统方法,检测正确率确有提升。Speech recognition technology is no longer limited to the laboratory.However,some key technologies still need to be further optimized and improved to adapt to the increasing demands of society and users.Starting from the technical theory of speech recognition and focusing on the endpoint detection technology,based on which,it is planned to improve the existing algorithm and enhance its recognition accuracy in a low signal-to-noise ratio environment.A new endpoint detection algorithm using cepstral distance-entropy ratio after LMS adaptive noise reduction is proposed,and the corresponding test is performed on the MATLAB platform.The results show that compared with traditional methods,the detection accuracy of this algorithm is indeed improved.

关 键 词：语音识别 自适应滤噪 倒谱距离 能熵比 

分 类 号：TN912.3[电子电信—通信与信息系统]