摘要
在基于内容的音乐研究中,正确提取音符起始点信息是识别音高、节拍、节奏、段落等音乐高级特征的基础.本文提出了基于匹配追踪(Matching Pursuit,简称MP)的两种新型音符起始点检测算法:基于MP解释程度和基于分音变化的检测算法.这两种算法均在MP分解的基础上,分析MP码本,并利用改进的峰值提取算法生成音符起始点向量.从实验结果看,本文提出算法的性能指标和MIREX 2011的最好结果相当.
Note onset detection is the preliminary work for recognition of high-level musical feature,such as pitch,rhythm, tempo and paragraphs,in content-based music information retrieval.This paper proposed two novel algorithms based on matching pursuit（MP）：the algorithm of MP degree of explaination and the algorithm of MP change of partial.Firstly,the musical signals were decomposed through MP,and then the code books were analyzed with the two algorithms.Finally,a modified peak-picking algorithm was applied to generate note onset vectors.The experiments showed the performance of our algorithms was nearly as good as that of 2011 MIREX.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2013年第6期1225-1230,共6页
Acta Electronica Sinica
基金
国家自然科学基金(No.60902065)
关键词
音符起始点检测
匹配追踪
MP解释程度
MP分音变化
note onset detection
matching pursuit
MP degree of explaination
MP change of partial
