Segmentation and Recognition using Artificial Neural Networks

The following is a proposal for separating cursive characters into separate and distinctive characters for recognition. The proposed method begins with thresholding of gray level image into binarised image using iterative thresholding selection method. The binary image is then slant corrected. Using Neuroheuristic technique, the slant corrected word is over segmented and artificial neural network trained with segmentation points is used to verify the segmentation points found and each segmented character is further extracted using character matrix extraction module. At last, using neocognitron simulator, the ANN is trained with the extracted character patterns and then, input pattern file containing a set of test patterns are tested for recognition on the basis of similarity in shape between patterns, but with a little effect from deformation, changes in size or shifts in position. Key-Words:Thresholding, Binarisation, Slant Correction, Neuro-heuristic Segmentation, Character matrix extraction, Artificial Neural Networks, Pattern Recognition.