A novel speech endpoint detection based on multiple complexities and fuzzy C means

Accurate Speech endpoint detection is important for speaker recognition, speech recognition, coding, and transmission and so on. In this paper, a fusion feature is proposed for speech endpoint detection, which utilized zero-crossing rate, Lempel and Ziv complexity (LZC), C 0 complexity and fluctuation complexity to represent the speech signal. In order to classify speech signal and background signal, the fuzzy c-means (FCM) is adopted as the classification. Experiments are carried out with white noise of NOISE-92 database to demonstrate the efficiency of the proposed method. Experimental results show that the proposed method can detect endpoints accurately.