Bangla Hand Written Character Recognition Using Support Vector Machine
Riasat Azim, Mohimenul Karim, Wahidur Rahman · Zenodo (CERN European Organization for Nuclear Research) · 2016
Recognizing handwritten character using computer is still consider a strong area of research. A fundamental problem in the field of Bangla character recognition is the lack of availability of Bangla handwritten character data set. In this thesis our main objective is to generate a larger dataset of Bangla character and as well as improving the recognition rate using Support Vector Machine. Support Vector Machines (SVM) is used for classification in pattern recognition widely. In our proposed method we applied support vector machine for increasing the recognition rate. A scanner is used to capture handwritten data sheet written in white paper by various people. After that several approaches used to generate the final data set for training and testing in SVM. A cropped image is scaled into 16*16 pixel matrix and then combing large number of image a dataset is produced. A binary classification technique of Support Vector Machine is implemented and rbf kernel function is used in SVM. This rbf SVM produces 93.43% overall recognition rate which is satisfactory result among all techniques applied on handwritten Bangla handwritten character recognition system.