作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2015, Vol. 41 ›› Issue (1): 186-189. doi: 10.3969/j.issn.1000-3428.2015.01.034

• 人工智能及识别技术 • 上一篇    下一篇

基于节奏和韵律调制谱特征的音乐流派分类

庄严,于凤芹   

  1. 江南大学物联网工程学院,江苏 无锡 214122
  • 收稿日期:2014-01-17 修回日期:2014-03-12 出版日期:2015-01-15 发布日期:2015-01-16
  • 作者简介:庄 严(1989-),男,硕士,主研方向:模式识别,信号处理;于凤芹,教授、博士。
  • 基金资助:
    国家自然科学基金资助项目(61075008)

Music Genre Classification Based on Modulation Spectrum Features of Rhythm and Rhyme

ZHUANG Yan,YU Fengqin   

  1. School of Internet of Things Engineering,Jiangnan University,Wuxi 214122,China
  • Received:2014-01-17 Revised:2014-03-12 Online:2015-01-15 Published:2015-01-16

摘要: 音乐主要包括形成节奏的冲击成分和形成韵律的和声成分,直接从音乐信号中提取特征会受到这2种成分相互影响。利用节奏与和声在时频平面具有不同规律的特点,通过对音乐信号进行谱图滤波,分离出音乐中的打击成分与和声成分。对打击与和声谱图分别进行小波调制,得到表现音乐节奏和韵律谱规律的调制谱特征,将其作为音乐流派分类中的长时特征。仿真实验结果表明,分离后的打击与和声成分谱图清晰地表征了音乐节奏和韵律的特点和规律;对8类音乐流派提取打击与和声调制谱特征,经线性鉴别分析降维后利用支持向量机进行分类,分类准确率达到73.54%。

关键词: 谱图分离, 中值滤波, 小波调制谱, 节奏和韵律, 中级特征, 音乐流派分类

Abstract: Music is mainly composed of percussive component and harmonic component,and the former forms the rhythm while the latter forms melody and harmony.Extracting features from the music samples directly are affected by the interaction between the two components.As the rhythm and harmony presenting different distributions in the time-frequency plane,the percussive component and the harmonic component can be separated by applying filtering on the spectrogram.It modulates the percussive and harmonic spectrograms with wavelet respectively and then gets the music rhythm and rhyme modulation spectrum features,which describes the long-term mid-level features of music genres.Experimental results show that the music rhythm and rhyme features represent feature and rule of rhythm and rhyme after the percussive and harmonic spectrogram separation.And the classification accuracy is 73.54% for eight music genres classification applying this method with Linear Discriminant Analysis(LDA) and afterward Support Vector Machine(SVM).

Key words: spectrogram separation, median filtering, wavelet modulation spectrum, rhythm and rhyme, intermediate feature, music genre classification

中图分类号: