Automatic Recognition of Handwritten Score Digits

Despite printed text being widely used since the introduction of computers and printers, several areas such as office automation, e-government, banking and education field still rely on manual data entry. Undeniably, manual data entry is very time consuming and human are prone to make mistakes during this task especially when the amount of the data to be entered is huge. Thus, recognition of handwritten digits plays an important role in life nowadays as it speeds up the data entry process. However, handwritten numerals recognition is a challenging problem as the handwriting styles are varying from person to person. In this project, a handwritten numerals recognition system is developed using Histogram of Oriented Gradients (HOG) as the feature extraction method. Several classifiers were also examined to determine the classifying method with the highest accuracy. The handwriting samples are scanned using an optical scanner and converted into digital images. After that, pre-processing steps such as segmentation, size normalization, and noise removal are applied to the scanned image to facilitate the feature extraction process.