Journal of Frontiers of Computer Science and Technology ›› 2011, Vol. 5 ›› Issue (1): 50-58.DOI: 10.3778/j.issn.1673-9418.2011.01.005

• 学术研究 • Previous Articles     Next Articles

Study on Multi-modal Music Genre Classification

ZHEN Chao, SONG Shuang, XU Jieping+   

  1. School of Information, Renmin University of China, Beijing 100872, China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-01-01 Published:2011-01-01
  • Contact: XU Jieping


甄 超, 宋 爽, 许洁萍+   

  1. 中国人民大学 信息学院, 北京 100872
  • 通讯作者: 许洁萍


According to the automatic music genre classification, propose a concept of multi-modal music genre classification. For the feature selection step in the traditional method which is based on the low-level acoustic features,realize a novel feature selection algorithm interaction based forward feature selection (IBFFS). Use LDA(latent Dirichlet allocation) model tags which are available on the Internet, convert the probability of tags for each genre into the probability of music resources for each genre responding to the tags.

Key words: music genre classification, interaction based forward feature selection (IBFFS), feature selection, music tags, latent Dirichlet allocation (LDA)

摘要: 针对自动的音乐流派分类这一音乐信息检索领域的热点问题, 提出了多模态音乐流派分类的概念。针对传统的基于底层声学特征的音乐流派分类中的特征选择环节, 实现了一种全新的特征选择算法——基于特征间相互影响的前向特征选择算法(IBFFS)。开创性地使用LDA(latent Dirichlet allocation)模型处理音乐标签, 将标签属于每个流派的概率通过计算转换为对应的音乐属于每个流派的概率。

关键词: 音乐流派分类, 基于特征间相互影响的前向特征选择算法(IBFFS), 特征选择, 音乐标签, LDA 模型

CLC Number: