Research on Yunnan Folk Music Classification Based on the Features of HHT-MFCC

Folk music is an art form that expresses human emotions and carries rich deep meanings. The classification of folk music helps to understand the history and culture of the nation. Most of the speech signal processing is based on traditional Mel-frequency cepstral coefficient (MFCC) features. This method has the problem of single feature module, which causes problems such as poor accuracy of speech signal recognition. Targeting to study folk music, this paper based on Hilbert-huang transform (HHT) to extract the HHT-MFCC features of folk music, and combined MFCC and HHT-MFCC to obtain MFCC_HHT-MFCC. With three kinds of features, classification models of random forest and SVM are built to classify folk music. Experiment results show that with the MFCC_HHT-MFCC features, random forest can achieve a recognition rate of 92.59% and SVM can reach 92.39%, which is an improvement compared with MFCC and HHT-MFCC.