Automatic Tonic (SA) Detection Algorithm in Indian Classical Vocal Music

Abstract

Analysis of Indian classical singing performances requires notes to be detected from the extracted pitch profile of the songs. This in turn needs the knowledge of the tonic used by the singer. This is not directly obtainable and if the service of a musician is exploited the correctness would depend on the expertise of the musician. Also such a service may not be readily available. It therefore makes sense to put efforts on finding an algorithmic approach to the detection of the tonic (Sa). An expert musician can often identify the base tonic (Sa) in a singing. However how they do it is an inexplicit knowledge. It may, therefore, be presumed this to be some sort of error-feedback mechanism. Here in this paper we have presented an algorithm for the automatic extraction of the tonic (Sa) from the steady state pitch sequences of the actual song sung by the expert musicians. The data-base consist 118 aalap performances covering four ragas by 34 male and 7 female well known singers. As there is confusion on the actual ratio system used by Indian performers, we have tested in total 3 ratio system (12 note system) has been experimented with. The detected Sa were found to be in good agreement with the measured values.