A Hybrid Cluster-Classifier model for Carnatic Raga Classification

The two forms of Indian Classical Music - Hindustani and Carnatic, are both melody and rhythm-based systems as opposed to the harmony-based Western Classical Music. Raga in Indian Classical Music is a structure formed by different combinations of notes or frequencies. There are 72 main ragas in Carnatic music and there are many more derived ones out of each of these 72 ragas. In this work, a hybrid cluster-classifier machine learning model has been proposed to classify two Carnatic ragas. The ragas considered for validation are Sankarabharanam, Kalyani, Arabhi, Bilahari, Hamsadhwani, Mand, Shuddha Saveri, and Mohanam. It is shown that the proposed hybrid cluster-classifier model, using spectrum magnitude and phase features, performs better than the stand-alone classifier model. The average classification accuracy for the hybrid model is 0.76. The highlight of the proposed approach is that there is no need for preprocessing steps.