Raga identification in CARNATIC Music using hidden Markov model technique

Abstract

Raga identification is very essential for automatic Music Information Retrieval (MIR) systems. One of the methods of classifying and organizing the songs of South Indian Classical music (Carnatic Music) is by identifying the Raga used to compose the music. Ragas can be defined as melody types, method of organizing tunes based on certain natural principles. As there are thousands of ragas classes defined in carnatic music system, the process of identifying raga of a song is a difficult task. In this paper, we have proposed a method where we divide the raga classes into number of groups based on the jump sequence of swaras used in the raga. So, using this method the raga identification process is solved in two levels where in the first level we identify the group to which song belongs and then in the next step identify the raga. Here in this paper, we describe the method used to group the ragas and then present the experimental analysis of the work done. © 2012 Springer-Verlag.