Music Mood Classification Algorithm Considering Simulated Annealing Algorithm

The research of audio signal classification is very important to promote the development of automatic intelligent audio classification system. The research of audio signal classification, feature extraction, feature set optimization and classifier design are the most important three links. By studying the principle of the simulated annealing algorithm, the improved BP neural network is easy to fall into the local optimal defect, and the improved BP neural network classifier is used to carry out the simulation experiment. The experimental results show that the music classification accuracy of the proposed classifier reaches 87.8%, which is 2.6% higher than that of DNN, and its performance is obviously better than other network structures. The above results show that the performance of the music emotion classification algorithm applied in this paper is better than most of the algorithms in the market, and can make a certain contribution to the music recommendation system.