Research And Implementation Of Classification And Retrieval System Of Music Collection Based On Automatic Music Genre Classification

With the rapid and continuous develop of network and large database technology, today people could access online music database containing millions of music piece. These music belongs to so many different genres. That how to manage these music collection effectively to offer the function to help people to access and retrieve music collection that them don't know to find what them like,is becoming a importance problem. The common method is to use the significant information, for example genres,names of musician, areas, the instrument. But these informations may be not complete, and complete them with suggestions of listeners will be a huge work that impossible to be done. A more developed method is to use signal and mode recognition technology to analyses the music file by automatic computer dispose to get significant information. automatic genre classification is more developed amout these technologies.This paper expound the main technologies of this field. This kind of technology consisted of two steps. The first step is to attain digital feature expressing genre character, the second step is to classify the feature by model recognition and machine learning technology. The music analyses technology consists two steps, too. The first step is to segment the music piece into very short time piece, and by analyses the spectral of the wave we can get the timbre character of the music. The second step is to static and analyses the timbre in the scale of whole file, by doing this we attain the music genre feather. The classification technology is consisted of two kinds—the supervised method and the unsupervised method, the supervised used training data set with class label and the unsupervised is not need training set but the accuracy will be lower.Our study is based on music genre classification technology. The system of managing and classification huge music collection is be proposed. It is consisted of three modules that is music analysis module, classification module, visualization module and UI. It can classify music pieces into different genres to be organized effectively. User can also type in genre names to retrieve the music or present music genre similarity relationship on a plane, by transforming auditory sense into visualization sense, people can explore the music collection intuitionally. And the system performance could be developed by user feedback to suit people's interest.