Abstract: The traditional method of short time Average Magnitude Difference Function（AMDF） often appears the mean downward trend that leads to finding the valleys which are not global lowest point. It also occurs the halving frequency and the doubling frequency errors in pitch tracking. To resolve this problem and enhance the valley value features, an improved algorithm based on AMDF is proposed in this paper. Firstly, the traditional AMDF is decomposed by using Empirical Mode Decomposition（EMD） and reconstructed after removing the trend component to obtain new EMD-AMDF. Secondly, the improved weighted feature of EMDAMDF/ACF is extracted to detect the speech pitch. Finally, experimental results show that this method can increase the valley features of AMDF and improve the accuracy of pitch detection.

Key words: pitch detection, Average Magnitude Difference Function（AMDF）, Empirical Mode Decomposition（EMD）, Auto Correlation Function（ACF）

摘要： 针对传统的短时平均幅度差函数（AMDF）法由于出现均值下降趋势，谷点并非全局最低谷点而导致基音周期提取中的倍频和半频错误出现的情况，提出一种改进算法。将传统的AMDF经过经验模式分解（EMD）处理后去掉趋势项重组新的EMDAMDF，再利用短时自相关函数（ACF）对其进行加权，构造新的EMDAMDF/ACF加权平方特征检测该语音帧的基音。仿真结果表明，该方法有效地加强了AMDF的谷值特性提高了基音周期检测的准确率。

关键词: 基音周期, 平均幅度差函数, 经验模式分解, 自相关函数