Old Handwritten Music Symbol Recognition using the Combination of Foreground and Background Projection Profiles

Optical music symbol recognition got huge attention during the last decade, lot of works have been reported for printed music symbol recognition, but very limited works have been reported for old handwritten music symbol recognition. Degradation due to aging, noise, change in handwriting styles and size makes this problem more challenging and difficult. In this paper, we have attempted the problem of old handwritten music symbol recognition. To do this, we applied three basic steps such as pre-processing, feature extraction, and classification. In pre-processing, each isolated character is resized to the size of 64×64. The projection profiles are computed from foreground as well as from background in feature extraction. The task of classification is carried out using simple k-NN classifier. Promising results were achieved with the accuracy of 95.72%.