Research on Music Emotion Recognition and Intelligent Classification System Based on Deep Learning

Aiming at the problem of music emotion recognition and classification, this paper proposes an intelligent system based on deep learning. Firstly, we introduce the theoretical basis of music emotion and its application in music information retrieval. Secondly, the deep learning models adopted, including time convolutional neural network (TCN) and long short-term memory network (LSTM), are elaborated, and how to use these models to process the time-frequency characteristics of music signals to achieve emotion classification. In addition, we employ attention mechanisms to enhance the model's ability to recognize key emotional features in music. In the experimental part, this paper uses a public music emotion database for model training and testing, compares the performance of different deep learning models in music emotion recognition tasks, and analyzes the relationship between model performance and music characteristics. The results show that the model combined with TCN and LSTM performs well on the recognition task of multiple emotional states. In addition, the accuracy and robustness of emotion recognition improved significantly after the addition of attention mechanism.