Handwritten Arabic numerals recognition system using probabilistic neural networks

Abstract

This paper presents a system for the recognition of the handwritten Arabic numerals zero to nine (0–9) using a probabilistic neural network (PNN) approach. This system can recognize handwritten input and externally imported Arabic numerals in real time, including two processes of image pre-processing and recognition. Image pre-processing involved normalization and expansion to enlarge the characteristics of the picture for easy recognition. Recognition process involves the calculation of mode distance, which can help obtain the similarity match between sample matrix and learning matrix. According to similarity matching the input feature matrix will be classified into one of the ten numbers. This study uses MATLAB to establish a graphical interface that is easy to operate. The system has good performance and strong scalability, which established a simple experimental platform for a more in-depth study of the recognition of handwritten Arabic numerals.