关键词: 语音端点检测, 最短时间门限, 短时高脉冲噪声

Abstract:

Endpoint detection of speech signals is crucial in speech recognition.Accurate endpoint detection can reduce the amount of computation of subsequent processing of speech signals and save system resources.Now many methods of the detection,such as instantaneous energy frequency value,spectral entropy and wavelet energy entropy,can accurately and effectively detect voice segments.This paper introduces a improved method for dual-threshold endpoint detection using short-time average zero-crossing rate and short-time average energy,then seting up a shortest timely restricted value to effectively detect Chinese speech in the high-pulse noise condition.A comparison with other methods shows that our method can effectively improve the recognition rate with high-pulse noise.Simulation results show that an accuracy of as high as 93% for endpoint detection can be achieved.

Key words: endpoint detection of speech signals;shortest timely restricted values;short-time and high-pulse noise