Abstract:
Speech endpoint detection is an important part of speech processing and recognition.Traditional endpoint detection methods have low accuracy in speech endpoint detection and poor anti-noise ability under the condition of low signal-to-noise ratio, In this paper, an improved zero-entropy feature parameter speech endpoint detection algorithm is proposed.This method studies the short-time zero-crossing rate, short-time energy and basic spectral entropy of three speech endpoint detection feature parameters, and proposes a new speech parameter, namely, the short-time energy zero-entropy value. Finally, a two-threshold algorithm is adopted to carry out endpoint detection.Simulation results show that compared with the traditional zero-energy ratio endpoint detection method, this method has higher endpoint detection accuracy under different low SNR conditions.