The Study And Application On Technology Of Music Mood Classification System

With the rapid development of informalization in our country, a variety of multi-media materials have been produced rapidly. A series of new technology of automatic indexing and searching on multi-media materials is required to realize effective storage, communication and share of increasing materials. This paper is an in-depth study of the music mood automatic classification technology.All of the existing music searching systems are primarily limited to the application of Text Searching technology, while the input of information still mainly depends on handwork. It is difficult for these systems to effectively index and search the rapidly increasing music materials. On the one hand, it is increasingly hard for the information based only on text to satisfy the searching requirement of users. On the other hand, it is impossible to deal with so many musical works in time only by handwork. Therefore, a new challenge for indexing and searching technology on how to meet these requirements by adopting new technology comes into being.For the above-mentioned new characters of music indexing and searching technologies, a music mood classification searching technology based on content is put forward in this paper. By analyzing and comparing the existing music mood classification technologies based on content, an automatic classification system of music mood is designed in this paper, in accordance with the technical requirement of national Radio&TV industry and the system structure and technology of GMM music mood classification. What's more, through the conformation of fitness of all kinds of audio characters in classification progress after a great deal of examination and analysis, an improved level method of classifier focusing on the unreasonable unified of the existing classifiers is proposed, to some extent, which can improve the accuracy of classify systems. In addition, the pre-segment of music based on mood is also studied. In this paper, the K-L distance is adopted to compute the difference between adjoining units, which divides a whole song into several segments each with the same mood so that a solid foundation is laid for better classification of music mood.The automatic classification and searching technology of music mood is realized in the Audio Asset Management System of Shanghai Media Group with a very good outcome, accumulating precious data for the further analysis of music mood.