A Multipitch Approach to Tonic Identification in Indian Classical Music

The tonic is a fundamental concept in Indian classical music since it constitutes the base pitch from which a lead performer constructs the melodies, and accompanying instruments use it for tuning. This makes tonic identification an essential first step for most automatic analyses of Indian classical music, such as intonation and melodic analysis, and raga recognition. In this paper we address the task of automatic tonic identification. Unlike approaches that identify the tonic from a single predominant pitch track, here we propose a method based on a multipitch analysis of the audio. We use a multipitch representation to construct a pitch histogram of the audio excerpt, out of which the tonic is identified. Rather than manually define a template, we employ a classification approach to automatically learn a set of rules for selecting the tonic. The proposed method returns not only the pitch class of the tonic but also the precise octave in which it is played. We evaluate the approach on a large collection of Carnatic and Hindustani music, obtaining an identification accuracy of 93%. We also discuss the types of errors made by our proposed method, as well as the challenges in generating ground truth annotations.