基于多音轨聚类的MIDI音乐主旋律提取方法

Melody Extraction Method from MIDI Based on Polyphonic Clustering

MIDI是音乐数字接口,可以称为计算机可以理解的乐谱,而一首乐曲中最终要的部分就是主旋律部分,它包含了乐曲中最重要的信息。主旋律提取的目的是可以将不同风格的乐曲进行更准确地分类。论文将选择MIDI文件,采用多音轨聚类算法,提取乐曲中的主旋律。首先将各个音轨块中其他伴奏信息删除,求出个音轨块的平均音高向量,接着对音轨块进行聚类和筛选,最后在筛选后的音轨块中提取出主旋律。通过与标准主旋律的对比,该算法可以有效地从乐曲中提取出主旋律,提高了准确性,为乐曲风格的分类提供了基础。

MIDI is a digital interface for music,which can be called a score that a computer can understand,and the final part of a piece of music is the theme part,which contains the most important information in the piece.The purpose of theme extrac-tion is to classify different styles of music more accurately.In this paper,MIDI files will be selected and multi-track clustering algo-rithm will be used to extract the main melody of the music.Firstly,other accompaniment information in each track block is deleted to find the average pitch vector of each track block,then the track block is clustered and screened,and finally the main melody is extracted from the screened track block.By comparing with the standard theme,the algorithm can effectively extract the theme from the music,improve the accuracy,and provide a basis for the classification of music style.