Emotion Recognition of Chinese Traditional Folk Music using an Assembling Machine Learning Method

Various papers published recently about the emotion ofwestern pop music, none have looked into how to describe Chinese traditional folk music. The accuracy of existing algorithms in recognizing emotions in Chinese traditional folk music is just 42%. The fundamental problem is the composing style of Western and Chinese traditional folk music is vastly different. As a result, we quantify the sentiment of music using "Enhanced XGBoost" Chinese traditional folk music model which integrates three judgment models: logistic regression, SVM, and XGBoost. For training, we use collected data sets ofWestern pure music and Chinese traditional folk music to determine four characteristics of Chinese traditional music, including pitch, loudness, rhythm, and timbre. The enhanced model's accuracy achieves 85 percent when recognizing the emotion of traditional Chinese folk music. We aims to provide technical assistance for the development of Chinese traditional folk music. The quantitative study of emotions can be used to help the platform produce personalized folk music recommendations for users.