Optimized handwritten character recognition using artificial neural network

Abstract

Handwritten character recognition (HCR) plays important role in the modern world and is one of the focused area of research in the field of image processing and pattern recognition. Handwritten character recognition refers to the process of conversion of hand-written character into printed/word file character which can immensely improve the interface between man and machine in numerous application. It is difficult to process with great variations in writing styles, different size and orientation angle of the character that are existing. Also segmentation of cursive handwritten text is difficult as the edges can’t be detected easily. There are numerous approaches to recognize handwritten data. The images are acquired using a digital camera or scanner and stored in standard format like JPG, PNG etc. The second stages include pre-processing techniques like Binarization, Skeletonization, thinning, resizing the image and segmentation. In our work we mainly concentrated on extracting statistical features of alphabets like mean, variance, standard deviation, Skewness and kurtosis, which differentiates a character from another. We used feed forward algorithm to train Artificial Neural Network (ANN). The features of input character after pre-processing are fed into ANN. A database of 650 samples is created to test input samples for recognition of character by neural net-work. The Experimental results that we have achieved show 88.46 % accuracy rate with minimum time taken for training.