语音信号端点检测的程序实现

Program Realization for End--point Detection of a Speech Signal

在孤立字识别中,精确地判别语言信号的起始点和终止点是相当重要的。确定出语音信号范围的方案可以用来减少大量非实时系统的计算和提高识别精确度。本文在利用语音的某些特征参数——短时平均幅度或能量和短时平均过零率的基础上,提出了利用上述特征参数进行语音端点检测的IBM/PC机实现程序。

In recognition of isolated words, it is essential to locate the beginningand the end of a speech signal precisely. A schemc for locating the regions of aspeech signal can be used to eliminate significant computation in nonreal--timesystems and improve the recognition accuracy. In this paper, a realized program forIBM PC (personal computer) is proposed to locate the end-point of a speech signal,which is based on the using some characteristic parameters of a speech signal--short--time average magnitude (or short-time energy) and short-time average zero-crossingrate.