Optical music recognition for traditional Thai sheet music

Abstract

Optical music recognition (OMR) is a system to transform a sheet music into a format readable by a machine. Over the last few years, several methods of OMR have been proposed for a standard sheet music. However, these methods cannot be applied directly for a traditional Thai sheet music (TSM) which uses Thai characters to represent music notes in the traditional way. This paper proposes a novel method to interpret an image of TSM into an editable or playable form. The proposed method contains three main processes including: 1) edge detection; 2) music note segmentation; and 3) music note recognition. First, canny edge detection is applied on an image of TSM. Second, music notes are individually segmented by using statistical analyses. Third, support vector machine (SVM) is used to recognize the segmented music notes. The experimental results based on many images of TSMs demonstrate that the proposed method can achieve very promising performance of above 80% accuracy, with a perfect performance (i.e. 100% accuracy) for some TSMs.