This work was done initally as part of the course Mini Project - CS350 as part of my Undergrad at the Department of Computer Science and Engineering at NITK Surathkal. Following this some more experiments were run and this work was submitted and accepted at the 6th International Conference on Machine Learning, Image Processing, Network Security and Data Sciences (MIND 2024) , which was hosted at NIT Goa. I had presented this work at NIT Goa and this is going to be a part of the Springer in Communications in Computer and Information Science (CCIS).
As a student of Carnatic Classical music, I have often struggled at times to identify the raga rendered in a song or to understand the nature of the raga rendered in a song. While a lot of work has been done with respect to Raga Identification, there is none done with respect to Classification into its various categories or types. On noticing this gap in literature, I thought of taking this up as my Mini Project, with the aim of publishing a paper out of this. While there are various types or classes into which ragas can be classified, the one type of classification is into Vivadhi and Avivadhi.
The dataset pertaining to this was collected from two sources namely - RagaSurabhi and YouTube. All the audio data was selected meticulously taking into consideration only songs which were rendered by a female in the same pitch without any instruments, with the exception of Tanpura. A total of 20 ragas with a split of 13 vivadhi and 7 avivadhi ragas.