摘要
端点检测是语音处理的关键技术。本文选取短时能量和短时平均过零率作为特征值,依据两级三门限端点检测的算法,其具体实现利用关系运算符,避免逐点比较循环本文语句,方便快捷地搜索到端点。经过Matlab编程仿真实验,结果表明,该实现方法有较高的准确性,对噪声具有一定的鲁棒性。
The endpoint detection is the key technology of speech signal process. Selecting the short time energy and the short time average zero crossing rate as the feature, according to the algorithm of two levels three-threshold detection, this paper uses relationship operator to avoid the point-by-point comparison looping statements, realizes quick and convenient search to endpoint. The restults of simulation experiments on the algorithm show that this method ires high accuracy, and makes certain robustness.
出处
《计算机与现代化》
2010年第3期188-190,共3页
Computer and Modernization
关键词
短时能量
短时平均过零率
门限
端点检测
short time energy
short time average zero crossing rate
threshold
endpoint detection
