In this paper, the forward neural network multi-feature fusion algorithm is used to extract the emotional features of music culture on artificial intelligence technology, considering the diversity and intermittency of the emotional features of the study, which needs to be parameterized. In the forward neural network architecture, the activation value obtained by using the nonlinear activation function is used, and the results obtained are passed to the next layer of data to realize layer-by-layer forward computation, which leads to the back-propagation activation function. The music culture emotion classification model is constructed based on the propagation mode of the forward neural network to determine the emotion recognition process. The research object is selected, the research process is determined, and in order to ensure the true validity of the research, it is necessary to test the reliability and validity of the research design scheme and to develop an empirical analysis of the comparison between popular music and traditional music culture. The results show that on the model, especially in the recognition of sacred, sad, passionate emotion type of music classification accuracy reached more than 88.2%. This paper’s model can improve the classification accuracy of music emotion to a certain extent. In the ontological knowledge analysis of popular music and traditional music culture, all three editions of textbooks show that general knowledge of music is predominant and has a large proportion, appreciation knowledge and extended knowledge are also considerable, and music knowledge is the least and has a small proportion. This study demonstrates the synergistic development of traditional culture and modern popular music, which is of great significance to the development of music education in colleges and universities.