基于统计模型的语音激活检测算法改进  

An improved voice activity detection algorithm of the statistical model-based

作　　者：张金榜 尹冬梅 

机构地区：[1]武警警官学院信息工程系,四川成都610213 [2]武警江苏省总队通信站修理所,江苏南京210019

出　　处：《微型机与应用》2015年第12期14-16,共3页Microcomputer & Its Applications

摘　　要：语音激活检测技术是应用于语音侦听领域降低节点能耗的关键技术之一,其核心是语音激活检测算法。针对基于统计模型的语音激活检测算法仅采用当前语音帧的参数来判断有无语音信号而带来的误检率高的问题,提出用相邻语音帧逻辑与运算的方法对其进行改进。试验表明:改进后的算法在误检率上明显低于改进前,提高了语音信号检测的准确性,有效地降低了节点能耗。Voice activity detection algorithm is the core of voice activity detection technique, which is one of key techniques to reduce power of the voice interception node in voice detection field. An improved voice activity detection algorithm of the statistical model-based is proposed aiming at the problem of high inaccurate estimation for adopting the current voice frame to estimate voice occur or not only. It has been improved by the operation of logic ＇and＇ between the border upon voice frame. The emulational result shows that the improved algorithm has reduced inaccurate estimation ration than the unimproved and improved the accuracy, which can reduce the power effectively.

关 键 词：统计模型 语音激活检测 检测算法 

分 类 号：TP274.4[自动化与计算机技术—检测技术与自动化装置]