Automatic Music Emotion Classification Research

Music is a popular entertainment for pleasure in internet and human digital life. A great of variety music is increasing at an incredible rate. The need for clarifying. organization and management of this kind of information is extensively expanding.There are some research works that have exploited the approaches to organizing music in emotion cue. Music emotion classification is a popular topic in recent years. However, Due to the emotion information is personal perceptual, subjective and ambiguous, music affect classification is a challenging problem through a computational approach.In this paper, we adopt an accurate definition of music emotion which has been demonstrated by previous research. In term of this thesis, a kind of music emotion is presented by two factors: arousal and valenceThe focus on collecting and labeling music clips to different affect types depends on manual labeling test. The database for experiments on extracting emotion from music is established after 400 clips are symbolized.The music emotion presented ways is various. Describing the emotion in a scale means is the fundamental works of the music emotion classification. Fortunately,a open source system MARSYAS is accessible.A new method is attempted which is named Information Cell Mixture Models (ICMM) to automate the task of music emotion classification. This framework has potential application in both unsupervised concept learning and supervised classification learning. This framework is acceptable for music mood classification because emotion is a vague concept and has a cognitive structure. The application of ICMM is also suitable for music emotion classification.