Music recommendation system can help teachers and students to discover the interested learning content from the complicated learning information, and assist teachers and students to teach and learn better. This paper improves the traditional collaborative filtering recommendation algorithm by integrating the collaborative filtering based on user attributes on the basis of user-based collaborative filtering, and then combines the Canopy algorithm with the K-Means algorithm to improve the accuracy and computational efficiency of the recommendation algorithm, and then designs a personalized music recommendation system based on the recommendation algorithm. And it is applied to two music teaching classes in a university for aesthetic education, which proves that the method proposed in this paper has an important and extensive positive impact on enhancing students’ emotions in music learning. From the practical data of the control class and the experimental class, it can be found that the p-value of the topics in the pre- and post-test scores of learning interest in the experimental class are less than 0.05, which is a significant difference, indicating that the learning interest in aesthetic education in the experimental class has been significantly improved after applying the method proposed in this paper for music education.