Automatic classification of Indian Music using EMD

Abstract Indian music is very rich in emotional content and technicalities. It is also contrasting having multiple genres which may be broadly classified as Classical (Hindustani or Carnatic) music, Semi-classical music (including Ghazals) and Light (including folk) music. The so called “rasa” or emotional content in Indian music is of much significance, e.g. selection of the right kind of music for therapeutic intervention by the music therapist who might be an expert in western art music (WAM) but could be new to Indian music. This paper presents a novel classification technique based on empirical mode decomposition (EMD), applied for the first time in Indian music. In this work, two genres of Indian music, Classical and Semi-classical, are considered. Highest classification accuracy of 98.68% is reported in the paper using SVM binary classifier after extraction of relevant features such as time domain features, frequency domain features, MFCC and Chroma vector.