Recognition Of Handwritten Word Images
Umesh D. Dixit, Rahul Hiraskar, Raghavendra Purohit, Sagar Shivanagutti · 2020 IEEE Bangalore Humanitarian Technology Conference (B-HTC) · 2020
Handwritten Character Recognition (HCR) is one of the fields in which a lot of research work is going on. In recent years it has taken quite a significant place in the research area. The objective of HCR technology is to transcript the handwritten scripts to digital form which is much more advantageous in storing than manual scripts. This paper presents a method for hand written word image recognition, which can be later extended for entire document. The proposed system employs Histogram of Oriented Gradients (HOG) features of the character images to train the classifier. The trained classifier is then used to recognize the segmented characters of word image to form a computer readable word. This work also compares the results using K-Nearest Neighbor (K-NN) and Support Vector Machine (SVM) classifiers. The SVM classifier is found to provide better word recognition rate of 75%.