摘要
语音端点检测是语音处理分析过程中的重要环节之一,该文介绍了语音端点检测的两个传统算法即短时平均能量和短时过零率,并将这两种算法结合起来进行藏语语音进行端点检测。运用Matlab编程和仿真验证了双门限判断法在藏语语音端点检测中的准确性。这种方法降低了藏语语音处理的时间、提高了处理的质量、可用来进行一些藏语语音识别的特征参数的提取。
Tibetan speech endpoint detection is an important part of Tibetan speech signal processing. This article introduced two traditional endpoint detection methods which including the short-term average energy and short-term zero rate algorithm and tested Tibetan speech endpoint detection by combing these two algorithm. It verified correctness of dual threshold method for the implementation of Tibetan speech endpoint detection.This method reduces the processing time and improve the quality for Tibetan speech Tibetan Speech Endpoint Detection.It can be used to extract some of the characteristics of Tibetan speech recognition parameters.
作者
卓嘎
边巴旺堆
姜军
ZHUO Ga, BIANBA Wang-dui, JIANG Jun （Electronic Information Department of Engineering School, Tibet University, Lhasa 850000,China）
出处
《电脑知识与技术》
2014年第11期7466-7469,共4页
Computer Knowledge and Technology
基金
基于国际编码标准的藏文词语排序算法研究（61163013）
关键词
藏语语音
短时平均能量
短时过零率
双门限
端点检测
Tibetan speech
short-term average energy
short-term zero cross rate
dual threshold
endpoint detection
