Research on Music Emotion Classification Based on Lyrics and Audio

With the development of audio recognition and research, more and more attention has been paid to the field of music emotion classification. However, it often can not get the expected effect of classification through parameter optimization and algorithm improvement of a single classification model to improve the accuracy of emotion classification, which is often limited by the characteristics of the model itself. Based on the techniques of text processing and music content in the computer field, the paper proposes combining lyrics classification with audio classification, so as to achieve the purpose of emotion classification. In terms of lyrics classification, LSI is used to reduce the dimensionality of the lyrics matrix, which effectively removes noise and uses SVM for classification. In terms of audio classification, BP neural network is used to complete the emotion classification of music; Finally, the two classification will be fused based on the improved algorithm of LFSM. The experimental results show that the combination of multiple classification methods for music emotion recognition can achieve higher accuracy of classification.