Research On Multi-Label Music Emotion Based On Evidence Theory And Information Cell Model

This paper focuses on multi-label music emotion recognition, models it and assesses the performance of recognition.This paper introduces the music emotional space and the multi-label music, and chooses an appropriate data set as the experimental data set. Then, we use Information Cell Model to produce the mass function, and finally use D-S evidence theory to integrate the multi-label information and achieve the ultimate identification of the music emotion. This paper gives out the evaluation criterion of the performance and the result of our experiment, and compares the result with not using information cells to compute the mass function, proves that using evidence theory to recognize multi-label music emotion combined with Information Cell Model to produce mass function has better results.This paper also experimentally proved that we can use Principal Component Analysis to reduce the dimension of music features when do the multi-label emotion recognition.This paper for the first time attempts to use Information Cell Model to give mass functions and integrate the multi-label evidences using evidence theory. Based on this method we also point out that Principal Component Analysis can be used to reduce the dimensions of music features.