Regression Approach for Shruthi Identification from Indian Classical Instrumental Polyphonic Audio

Shruthi for Indian classical music is similar to the tonic pitch of Western music. The key difference between Shruthi and tonic pitch is that tonic pitch has a fixed frequency for each note while shruthi is not constrained with limits. Knowledge of the tonic pitch is essential in Indian classical music in order to perform melodic analysis across artists and performances. The availability of a rich consumer base as well as large amounts of high-quality data has led the Music Information Retrieval (MIR) field to be biased toward Western music. The fixed value of frequency for a particular tonic makes it ideal to use a classification technique for its identification, while a regression-based approach is more suitable for Shruthi identification due to its unbounded nature. This work focuses on the less explored regression approach rather than the classification approach to identify the shruthi of a given piece of Indian classical polyphonic instrumental music.