类谐波积谱基音周期检测算法  

Pitch Detection of Harmonic Product Spectrum-like Function

作　　者：梅铁民[1] 付天娇[1] 朱向荣 

机构地区：[1]沈阳理工大学信息科学与工程学院,沈阳110159 [2]淄博民通热力有限公司,山东淄博255400

出　　处：《沈阳理工大学学报》2016年第2期14-17,23,共5页Journal of Shenyang Ligong University

摘　　要：基音周期检测是语音信号处理的一项重要内容。在低信噪比情况下,常用的自相关函数法容易出现半周期和倍周期估计错误。为解决这一问题,在谐波积谱法的启发下提出了基于自相关函数的类谐波积谱法。该方法不仅可以有效地克服谐波积谱法由于信号中某次谐波分量为零时所造成的困难,而且可以有效地提高低信噪比下基音周期检测的可靠性。Pitch detection is an important researching topic in signal processing. One of the frequently used algorithms is based on autocorrelation function,estimation errors frequently occur for low signal-to-noise ratio（ SNR） speech signals. To overcome such problems and motivated by the harmonic product spectrum（ HPS） method,a novel pitch estimation method is proposed by defining an HPS-like function of the autocorrelation of speech signal. The novel method can not only overcome the shortcomings of HPS if one of the harmonics has zero amplitude,but also be more robust under low SNR.

关 键 词：基音周期 自相关函数 类谐波积谱 

分 类 号：TN912[电子电信—通信与信息系统]