Chinese opera genre classification based on multi-feature fusion and extreme learning machine

Abstract

Chinese traditional opera plays an important role in Chinese traditional culture, it reflects the customs and value tendency of different areas. Though researchers have already gained some achievements, studies on this field are scarce and the existing achievements still need to be improved. This paper proposes a system based on multi-feature fusion and extreme learning machine (ELM) to classify Chinese traditional opera genre. Inspired by music genre classification, each aria is split into multiple segments. 19 features are then extracted and fused to generate the fusion feature. Finally, we use ELM and majority voting methods to determine the genre of the whole aria. The research data are 800 arias of 8 typical genres collected from Internet. This system achieves a mean classification accuracy of 92% among 8 famous Chinese traditional opera genres. The experimental results demonstrated that multi-feature fusion improves classification accuracy of Chinese traditional opera genres. Feature fusion is more effective than decision fusion in dealing with this problem.