Abstract
Raga are vital parts of Carnatic music and offer the performer a musical structure to improvise within them. Raga identification is considered to be a tedious task even for expert music listeners. As of now, we do not know a unique set of parameters which can precisely determine the raga of a Carnatic music. Even though several machine learning techniques are available to perform raga recognition, they all suffer from several issues like less recognition accuracy, poor scalability etc. In this work, we suggest a non machine model which has a high recognition accuracy. The model works on extracting five features of the music and then computing a similarity measure between the feature vectors of the input audio with the corresponding feature vectors stored in the database for raga recognition. Even though our system is not excellent in scalability, it is simple to implement and its recognition accuracy can be increased by increasing the feature vectors stored in the database.
Reference27 articles.
1. Rajeswari Sridhar and TV Geetha. Raga identification of Carnatic music for music information retrieval. International Journal of recent trends in Engineering, 1(1):571, 2009.
2. Rohith Joseph and Smitha Vinod. Carnatic raga recognition. Indian Journal of Science and Technology, 10(13):0974–6846, 2017.
3. V Kaimal and S Barde. Introduction to identification of raga in Carnatic music and its corresponding hindustani music. International Journal of Computer Sciences and Engineering, 6(6):955–958, 2018.
4. Devayani Hebbar and Vandana Jagtap. A comparison of audio preprocessing techniques and deep learning algorithms for raga recognition. arXiv preprint arXiv:2212.05335, 2022.
5. Dipti Joshi, Jyoti Pareek, and Pushkar Ambatkar. Indian classical raga identification using machine learning. In ISIC, pages 259–263, 2021.