DUPSO-DSKSVM: A Fast Music Classification Algorithm Based on Nonlinear Features

With the rapid development of online information, there has been an increase in the demand for music information retrieval among different types of music enthusiasts, making music classification an important research field. In this study, we propose a fast classification method for folk songs by combining the Dissipative Uniform Particle Swarm Optimization (DUPSO) algorithm and the Distance-based Sorting Kernel Support Vector Machine (DSKSVM) algorithm. This method leverages the nonlinear features of music to achieve accurate classification. DUPSO is used to optimize the penalty coefficients and kernel function parameters of KSVM, while DSKSVM is used to optimize the parameter optimization time of DUPSO. Experimental results show that the classification accuracy of the DUPSO-KSVM algorithm reaches 92%, and the parameter optimization time of the DUPSO-DSKSVM algorithm is only 17.38% of that of the DUPSO-KSVM algorithm, while still maintaining high classification accuracy.