Intelligent Classification Model of Oroqen Music Style Based on Deep Learning

This study applies deep learning toward building an intelligent identification model for the Oroqen music style, classifying, and recognizing the Oroqen traditional music style. A rich Oroqen music corpus was built by collecting existing audio materials. After pre-processing and feature extraction, these data are going to generate signal and Mel spectrograms, as well as MFCC characteristics, which are good to enter deep learning models. The Mel spectrogram is selected as the basis for this model, and the choice and design of the CNN architecture are followed; then, the whole process of data preprocessing, feature extraction, model training, and evaluation is combined to build the intelligent classification system. Experimental results showed that the proposed model performed well for most evaluation indicators, with a rate of 92% for accuracy, 91% for precision, and 90% for recall, while the harmonic F1 measure was 90%. It will therefore be able to effectively classify the music materials for Oroqen and archive them, so as to improve their culture and meanwhile be used as a reference by other studies regarding the ethnic music.