Music Emotion Recognition Model Integrating Deep Learning

Abstract

With the development of digital music technology, people begin to explore new classification and recognition methods to retrieve target music from massive data. Music is the carrier of emotion, and the recognition research based on music emotion classification has very important objective significance. The purpose of this paper is to study a music emotion recognition model incorporating deep learning. The audio features of music are extracted and screened based on the underlying audio features, fused with the audio features obtained by deep learning, and combined with the CNN-SVM network model for music emotion classification and recognition. The advantages of the two are combined to carry out the task of music emotion classification, and the final comparative experiments are carried out on three different datasets. Experiments show that the CNN-SVM model in this paper, combined with the filtering of the CNN layer and the new chord vector feature, achieves the best results on all three datasets.