改进的能量谱熵端点检测算法

Improved Endpoint Detection Algorithm Based on Energy Spectral Entropy

中文关键词:  短时能量  谱熵  端点检测  双门限检测  信噪比  正确率

英文关键词:short-term energy  spectral entropy  endpoint detection  double threshold detection  SNR  right accuracy

中文摘要:

      为了提高传统谱熵算法在信噪比较低环境下的端点检测效果，将短时能量特征与谱熵特征相结合，提出一种改进的能量谱熵特征，将模糊C均值聚类算法和贝叶斯信息准则结合对改进的能量谱熵特征门限进行估计，最后采用双门限算法进行端点检测。仿真实验结果表明，在信噪比为-5 dB的白噪声环境下，改进的能量谱熵算法的端点检测正确率为76.9%，远高于短时能量算法和谱熵算法，在低信噪比环境下具有更优的端点检测效果与稳健性。

英文摘要:

      In order to improve the endpoint detection effect of traditional spectral entropy algorithm in low signal noise ratio environment,combining the characteristics of short time energy and spectral entropy,a new feature of improved energy spectral entropy is put forward,and the threshold is estimated by the fuzzy C-means clustering and Bayesian information criterion algorithm.Finally,double threshold algorithm is used to achieve endpoint detection.The simulation results show that the accuracy of improved energy spectral entropy is 769% under -5 dB white noise environment,which is much better than the short term energy and spectrum entropy algorithms.The improved energy spectral entropy algorithm has better effect of endpoint detection and robustness in a low SNR environment.