Research On The Musical Genre And Instruments Classification Based On Sparse Representation-based Classification Via L～1-minimization

In recent years, with the development of the internet and digital music, the content-based musical research has become the hot spots of the research. The musical information retrieval is one important part of the content-based musical research. Recently, with the number of the music increasing gradually, we retrieval the music we need on the internet become very difficult. However,most of the musical classification methods are manually, with the development of the digital music, manual classification could not be deal with this development situation, so the development of the content-based musical classification becomes urgent and important.The study is based on the existing research. In feature extraction, this paper introduces the new timbre feature which mainly consist statistical features. A new classification method used in this paper’s audio classification system. The introduction of the new feature has been improved effective in the musical genre classification, which improves the accuracy of the classification. This paper compared the support vector machine and sparse representation-based classification, and the classification based on the sparse performs better than support vector machine. The experiment shows the combination the PCA and the classification method to improve the classification accuracy.Chinese classical musical instruments are an important part of the world musical instrument, and China has a long history of classical musical instruments. However, very little work study Chinese classical instrument, in this paper, the author research on Chinese classical instrument classification. This paper achieves high classification accuracy with the combination of the features and classification method.Through the classification of musical genres and instruments, the author illustrates the effectiveness of features introduced in this paper, and shows the advantages of sparse representation-based classification comparing to support vector machines.