Analysis on handwritten Bangla character recognition using ANN
Arifur Rahaman, Md. Mehedi Hasan, Md. Faisal Shuvo, Md. Abu Saleh Ovi, Md. Mostafizur Rahman · 2014
Bangla handwritten character recognition is one of the complex works because of the wide variation of the Bangla character. In this paper we proposed a new approach for extracting the features of Bangla handwritten characters and then recognition of those characters using artificial neural network has done. For the feature extraction process we have used a row and column basis segmentation process in which a binary image the ratio of the dispersion of black from the top of the segment and total number of column or row pixels is measured on that particular segment as well as for each segment and then from this data a feature matrix of M×1 dimension is created for each character. When features from all the characters are extracted then some of the data is used to train the feed forward neural network and after that some characters are used for testing. In our proposed method 94.3% recognition accuracy is obtained when tested on some handwritten Bangla characters.