Optical Music Recognition: Conversion of sheet music to Sargam

Introduction

This project was born out of a personal need to transcribe sheet music from western music notation (using alphabets A, A#, B, C... and others) into Indian sargam ( Sa (सा), Re (रे) and the others). However this is tricky on many levels -

• In order to convert the notations, we first need to get the western notations in the first place from the music sheets. Music sheets are documents which represent music using complex symbolic grammar, using an array of symbols for notes, time signatures, key signatures, clefs, silences, movements and everything else. Even then, if we change just one set of symbols (the key signature for example), while keeping everything else same, the symbols representing the notes will now represent completely different notes!! (and the music will sound weird too). This is not even considering the other technical challenges of dealing with pdf documents and images and what not.
• Given that somehow magically we have a one-to-one translation of a music sheet, again we can’t freely convert the notes into sargam. This is because we will need an acute idea of the bars, measures and the length of each note - without which we will end up with only a string of letters and no rhythm.

Given this myriad sinkhole of problems, this project aims to try and solve as much of these stepwise as possible.