哼唱检索中一种新颖的基音周期提取算法  

Novel method of pitch period extraction for QBH system

作　　者：丁娜娜[1] 潘峰[1,2] 杨晓元[1,2] 吴翔[1] 

机构地区：[1]武警工程学院电子技术系网络与信息安全武警部队重点实验室,西安710086 [2]西安电子科技大学网络信息安全教育部重点实验室,西安710071

出　　处：《计算机工程与应用》2009年第35期142-144,共3页Computer Engineering and Applications

基　　金：国家自然科学基金No.60573032~~

摘　　要：在哼唱检索(QBH)研究中,针对乐曲旋律表达方式以及有效的检索匹配算法已有一定的研究成果,但提取哼唱音频信息基音周期的算法研究还不充分。正确提取基音周期,降低哼唱信号基音周期提取过程中引入的误差,是有效提高整个检索性能的关键所在。在充分分析已有的基音周期提取方法基础上,提出了一种短时自相关(ACF)与短时平均幅度差(AMDF)函数相结合的CAA算法,仿真实验证明了新方法的有效性,相对于传统时域算法提取基音周期,准确度有显著提高。In the course of the research on Query By Humming（QBH）,music melody expressing method and effective retrieval algorithm have already had certain research findings,but the study based on the extraction of pitch period in audio information is not sufficient.Extracting the correct pitch,which can decrease the error in the process of extracting the humming audio signal,is the key to improve the whole QBH system s capability.On basis of the full analysis of the pitch extraction method that have been proposed,the article proposes a new CAA algorithm which combines Auto-correlation Function （ACF） with Average Magnitude Difference Function（ AMDF ）.Experimental results show the effectiveness of the new algorithm.Compared with the traditional time-domain pitch extraction algorithms,accuracy has improved significantly.

关 键 词：哼唱检索 基音周期 短时自相关 短时平均幅度差 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]