Indonesian's Traditional Music Clustering Based on Audio Features

Cluster analysis has been used widely in some applications. In this research, 101 songs from 18 provinces in Indonesia clustered by some set of features extracted directly from the audio data. Before the clustering process, feature selection process with PCA method performed using 60 audio segments from 4 songs to find the optimal set of features which will be used in clustering process. In clustering process, the selected features are extracted from audio signal and clustered by x-Means algorithm to find the proper number of cluster. Clustering with this method resulted 4 clusters. The result of this process shows the characteristic of each cluster and some distributions of cultures between areas and provinces. An Agglomerative Hierarchical Clustering method also conducted to compare the result.