Silence/Voice Segment Detection of Speech Signals Based on Bark Wavelet Transform Parameters

A silence/voice detection method for speech signals using Bark wavelet transform parameters is proposed.In this method,the ability of frequency segmentation and energy focusing in Bark wavelet is utilized to extract statistic parameters of speech signals in different subbands.By introducing parameter validity analysis based on fuzzy entropy,the most discriminable and stable parameter in all subband parameters is derived as discriminate parameter.Simulation analysis demonstrates the stability and effectiveness of this method under different noise conditions,and its accuracy and robustness are improved to a certain extent,compare to traditional parameters.