Recognition of Arabic numerals with grouping and ungrouping using back propagation neural network

Abstract

In this paper, the authors propose a method to recognize Arabic numerals using back propagation neural network. Arabic numerals are the ten digits that were descended from the Indian numeral system. Although the pattern of 0-9 is the same as in Indian numeral system, the glyphs vary for each numeral. The proposed method includes preprocessing of digitized handwritten image, training of BPNN and recognition phases. As a first step, the number of digits to be recognized is selected. The selected numerals are preprocessed for removal of noise and binarization. Separation process separates the numerals. Labelling, segmentation and normalization operations are performed for each of the separated numerals. The recognition phase recognizes the numerals accurately. The proposed method is implemented with Matlab coding. Sample handwritten images are tested with the proposed method and the results are plotted. With this method, the training performance rate was 99.4%. The accuracy value is calculated based on receiver operating characteristics and the confusion matrix. The value is calculated for each node in the network. The final result shows that the proposed method provides an recognition accuracy of more than 96%.