摘要
语音端点检测是语音处理中重要的领域之一。常规谱熵语音端点检测算法是通过检测语音的功率谱的平坦程度,从而达到语音端点检测的目的。但是该方法在平稳噪声环境下较好,在无噪声和非平稳噪声环境下效果较差。作者在分析了无噪声环境下常规谱熵端点检测算法效果差的原因的基础上,结合了语音的短时能量算法,对常规谱熵算法进行了改进,形成了一个新的特征参数——谱熵能量积。仿真结果显示,该方法相对于常规谱熵算法,在无噪声的环境下检测精度有了很大的提高,在非平稳噪声环境下也有了一定的提高,鲁棒性得到增强。
Robust endpoint detection is one of the most important areas of speech processing.The traditional spectral entropy algorithm used to speech endpoint detection is based on the flatness of the detected speech spectrum,but this method is effective in stationary noise environment rather than other noise cases.This paper analyzed the reason why the traditional spectral entropy algorithm was not effective in the situation without noise,at the same time,it showed a new characteristic parameter based on short-time energy and spectral entropy algorithm.The results show that this method has better robustness and precision,especially in no noise environment.
出处
《武汉理工大学学报》
CAS
CSCD
北大核心
2013年第7期134-139,共6页
Journal of Wuhan University of Technology
基金
中央高校基本科研业务费专项基金(2011-Ia-005)
关键词
端点检测
谱熵
短时能量
鲁棒性
endpoint detection
spectral entropy
short-time energy
robust
