语音信号的基音周期在语音压缩编码、语音分析与合成、语音识别和说话者确认等方面是很重要的参数。到目前为止,还没有一种方法能在不同的话者、不同的环境下都达到最佳的效果。根据汉语语音信号的特点,本文首先对语音进行滤波处理、中心削波处理来滤除语音中高次谐波成分及高频噪声,接着利用修正自相关函数提取语音基音周期,最后对所得的结果进行平滑处理。实验结果显示改进算法提高了基音提取的可靠性,最后本文在改进算法的基础上进行了声调识别。 The pitch period of a speech signal is a very important parameter in speech compression coding, speech analysis and synthesis, speech recognition and speaker verification. So far, there is no way to achieve the best results in different speakers and environments. According to the characteristics of the Chinese speech signal, this paper first filters the speech, the center clipping process to filter the high harmonic components of the speech and the high frequency noise, and then use the modified autocorrelation function to extract the speech pitch cycle, and finally the results Smoothing. The experimental results show that the improved algorithm improves the reliability of pitch extraction. Finally, this paper makes a tone recognition based on the improved algorithm.