Publications

Automatic music genre classification are essential for applications that provide music analysis and music retrieval as well as for recommendation systems. Most of the prior research in this domain focused on investigating western music genre classification. The automatic classification of Arabic cultural style music has never been studied. In this paper, we present an approach for classifying digital Arabic songs automatically based on their cultural style. Four Arabic cultural styles are studied which are the Moroccan, Egyptian, Shami and Khaliji. Three sets of acoustic features are investigated together with supervised classifiers to identify the cultural style of a song. Moreover, feature selection algorithms are employed to identify the most suitable subset of features for classification. An overall accuracy of 80.25% is reached for the classification of the four styles using a Decision Tree classifier after applying the OneR Attribute feature selection algorithm.