摘要
通过分析话音信号的时域特征,发现其具有短时能量对浊音敏感、短时平均过零率对清音敏感的特点,可以准确判断有声段和静音段。在此基础上提出以静音能量阈值和静音时延阈值作为双门限判定参数的话音信号端点检测算法。实验表明,该方法简单、准确率高、鲁棒性好。
By analyzing the time domain features of voice signals,it is found that the short-time energy is sensitive to voiced sounds,and the short-term average zero-crossing rate is sensitive to unvoiced sounds,and sound segments and muted segments can be accurately judged. Based on this,a voice signal endpoint detection algorithm is proposed,which uses the silence energy threshold and the silence delay threshold as dual threshold decision parameters.Experiments show that the method is simple,accurate and robust.
作者
郑璐
ZHENG Lu(College of Information Engineering,Longdong University,Qingyang 745000,Gans)
出处
《陇东学院学报》
2018年第5期4-8,共5页
Journal of Longdong University
关键词
端点检测
短时特征
双阈值
endpoint detection
short-term characteristics
dual-threshold
