A study on the classification of Yunnan folk songs based on ensemble learning

ABSTRACT:   Nowadays, music classification has become an important application field for classification algorithms. There is only little research about folk songs classification. This paper used the original folk songs in Yunnan Province as research objects, extracting their CELP audio characters on a 11-dimensional scale, comparing and analyzing the classified experimental data from different proportions of marked training samples, which are processed by a single classifier and an ensemble classifier (such as Bagging, AdaBoost, and MCS). Among them, AdaBoost is the most efficient one with an accuracy rate of higher than 85%. The result showed that the ensemble classifiers could mean the effectiveness of the study methodology.