K-means clustering analysis of Chinese traditional folk music based on midi music textualization
Zhang Liumei, Fanzhi Jiang, Jiao Li, Ma Gang, Liu Tianshi · 2021
The current mainstream feature extraction of music information retrieval (MIR) is based on acoustics, such as frequency, loudness, zero-crossing rate. while it is rare to perform feature extraction and music analysis directly on symbolic music. This article seeks to introduce the idea of text clustering in natural language processing into the field of symbolic music style analysis. From this, this work got inspiration to turn midi music into text data and transform it into weighted structured data through tf-idf, and then use the K-Means clustering algorithm to perform cluster analysis and comparison on the traditional Chinese folk music dataset we crawled, and finally use The T-SNE algorithm performs dimensionality reduction and visualization of high-dimensional data. After a series of objective indicators evaluation, it is proved that the clustering algorithm has achieved a good clustering effect on the midi note dataset we extracted; through the clustering results, comprehensive professional music theory knowledge and the historical development characteristics of traditional Chinese folk music, Reverse verification of the 1300 midi music data sets has distinct modal characteristics of traditional Chinese folk music.