Simulate Biological Neural Networks for Song Style Classification and Dynamic Resource Integration for Aesthetic Education

Xiaohui Yang
College of Big Data and Information Engineering, Baise Vocational College, Baise 533000, China.

Abstract:

Current audio classification methods cannot fully capture the time-varying characteristics of audio such as rhythm, melody, harmony, and the cold start and sparsity problems in the integration of aesthetic education resources. This paper designs a quantum-inspired heuristic algorithm based on simulated biological neural network (NN). In the framework of the algorithm, the Spiking Neural Networks (SNN) model is used to learn the temporal features of audio data to achieve accurate classification of song styles. In order to achieve dynamic integration of aesthetic education resources, the Neural Collaborative Filtering (NCF) model is used to generate personalized recommendations based on user behavior and the interactive information between aesthetic education resources. Based on these two goals, the Quantum Particle Swarm Optimization (QPSO) algorithm is used to simultaneously optimize the hyperparameters of SNN and NCF to improve the performance of both. The algorithm studied performed well in the song style classification task. Compared with mainstream models such as MusicNN, the classification accuracy (0.91), recall (0.93) and precision (0.92) all achieved the best results. In the test of integrating aesthetic education resources, the algorithm showed significant data adaptability: when the amount of user behavior data increased, the recommendation accuracy rate steadily increased from the initial 0.25 to 0.85, which was significantly better than the four comparison methods such as GNN. These experimental results show that the algorithm has important application value in the fields of music information processing and personalized recommendation, and its technical ideas can provide useful reference for the development of practical systems in related fields.