Machine Learning Based Indian Raga Classification and Detection

The detection of specific ragas, or melodic modes, within Carnatic music is a complex task that involves understanding and analyzing the unique melodic progressions characteristic of each raga. In this study, Hidden Markov Models (HMM) are employed as a computational tool to tackle this challenge. HMMs are utilized to learn and probabilistically represent the distinctive melodic transitions associated with each raga. Interestingly, the analysis focuses solely on the pitch contour of the music, indicating that even with this limited feature set, the HMM exhibits remarkable effectiveness in identifying ragas. Through experimentation with a diverse test dataset containing 16 samples spanning various modalities, the HMM-based approach achieves an impressive accuracy rate of 89.25%. This high accuracy underscores the robustness and adaptability of HMMs in capturing the intricate melodic structures inherent in Carnatic music. Overall, this study highlights the potential of computational techniques, particularly HMMs, in unraveling the complexities of musical patterns and contributing to the understanding and preservation of cultural musical traditions.