Bangla Handwritten Character Recognition Method

Abstract

Recognizing and extracting handwritten character information is still a challenge in the scanning process. This research describes a method for OCR applications where Bengali handwritten characters can be recognized effectively. This method mainly focuses on post processing steps like feature extraction and building classification model to find a favorable accuracy rate. For feature extraction LBP method is used. Local Binary Pattern is a coming of age feature extracting method which is applied very few times in Bengali Language. This work is also an experimental approach of confirming what occurs when LBP is used in Bengali Characters. To classify Random Forest algorithm is applied, which is also a unique classification method. The datasets are gathered by collecting Bengali characters written in various fashions. Initially, scanned images of Bengali characters are given as input and by applying LBP required features are extracted. Principal Component Analysis (PCA) is applied on the collected feature vectors to reduce the dimension. Finally, RF algorithm is implied on the output to generate a recognition rate. Support vector machine (SVM) is also used as a classifier to evaluate and compare.