OPTICAL RECOGNITION OF OLD HANDWRITTEN MUSIC SYMBOLS BASED ON TEXTURE DESCRIPTORS

Optical music symbol recognition facilitates to transcribe the music sheet into machine-readable format so that it can be used for various applications by converting it into midi format. Most of the works in the past have focused on the recognition of printed music symbols and a few on online music symbols. Earlier methods work very well for printed music symbol recognition. However, their performance is limited to clean and binarized documents. Handwritten music symbol recognition is explored a little as it has several challenges such as variation in writing styles, document degradation, noise etc. In this paper, we have investigated the performance of well-known texture descriptor namely Histogram of Oriented Gradients (HOG) for the Old Handwritten Music Symbol Recognition on the publicly available dataset. Support Vector Machine and K-Nearest Neighbor Classifiers were employed for the music symbol classification with K –Fold Cross Validation Technique. We have achieved encouraging results and shown the comparative analysis of various sizes of cell of computing HOG.