Electronic Music Type Identification Based on Multimodal Learning Optimized Residual Neural Networks

The classification effect of electronic music may be low due to the complexity of its type branches. Therefore, a research on electronic music type identification method based on multimodal learning optimization residual neural network is proposed. This paper analyzes the characteristics of electronic music from three perspectives of short-time energy, time-domain distribution and frequency-domain distribution, divides the original continuous music signal into several short segments (frames), and extracts the corresponding short-time energy, time-domain variance and frequency-domain variance information. In the stage of classification and recognition, the residual module is introduced into the neural network, and three characteristic modes are fitted to determine the final classification results. In the test results, the AUC of the recognition results of rhythmic electronic music and atmospheric electronic music is basically above 0.7, and the minimum value is also above 0.68. On the overall perspective, the AUG mean is 0.0783 and 0.0615 lower than that of the control group, respectively.