Abstract: The accurate pitch detection is difficult in speech processing. A new weighted pitch detection method based on Autocorrelation Function（ACF） is proposed. Utilizing that the Average Magnitude Difference Function（AMDF） has similar characteristics with the ACF, the ACF is weighted by the square of the reciprocal of the AMDF. The pitch contour is smoothed. Experimental results show that the algorithm improves the accuracy of pitch detection.

Key words: pitch frequency detection, Autocorrelation Function（ACF）, Average Magnitude Difference Function（AMDF）, pitch contour smoothing

摘要： 语音基音频率的准确检测是语音信号处理的难点之一。提出一种加权短时自相关函数（Autocorrelation Function，ACF）算法提取基音频率。在传统的ACF方法基础上，利用短时平均幅度差函数（Average Magnitude Difference Function，AMDF）的平方对ACF函数进行加权，由此加强短时自相关函数在基音周期倍数处的峰值特性。对提取出的基频曲线做平滑处理。实验结果表明，该方法提高了基音周期检测的准确率。

关键词: 基音频率检测, 加权短时自相关函数, 短时平均幅度差函数, 基频曲线平滑