An extensible Optical Music Recognition system


David Bainbridge
Department of Computer Science, University of Waikato, Hamilton, New Zealand.
d.bainbridge@cs.waikato.ac.nz

Tim Bell
Department of Computer Science, University of Canterbury, Christchurch, New Zealand
tim@cosc.canterbury.ac.nz


Abstract

Optical music recognition (OMR) is a form of structured document image analysis where symbols overlaid on the conventional five-line stave are isolated and identified so that the music can be played through a MIDI system, or edited in a music publishing system. Traditionally OMR systems have had recognition techniques hard-coded in software. This paper describes a system that has been designed to be extensible without the need to change the system's source code. Extensibility is achieved by providing tools for music recognition that are used to tailor the system to suit the type of music being recognised. The tools include a selection of methods for identifying staves and isolating objects from them, methods for describing and identifying primitive musical shapes, and a grammar for specifying the relationships between the shapes that are recognised. The system is flexible enough to work with different publishers' symbol sets, and even with different types of music notation, such as the square-note notation used in early music.
Conference Home Page