Hybrid GMDH model for handwritten character recognition

Character recognition has been a lucrative research area due to its application in various fields like human computer interaction, identification of human being or its personality through his handwritten characters. Its applications are increasing day by day. Also for a given language any alphanumeric character written by an individual is different and posts many computational challenges for its recognition. The main emphasis is on recognition cost, i.e. the accuracy and time required for recognition of character. Looking forward to advanced applications and to improve cost, a hybrid GMDH character recognition model has been proposed. Group Method of Data Handling concept is based on heuristic self organization in data mining where a statistical learning network is created using relationship between input and output variables. The input variables related to the output are retained in the model while those unnecessary get discarded. Hybrid GMDH is a statistical learning network were in relationship between input and output variables are represented as discrete entities at the initial layer and at the terminal layer, represented in terms of fuzzy set. This approach based on combining polynomial GMDH and fuzzy GMDH, named hybrid GMDH, is viewed to be more efficient to optimize recognition ability for handwritten characters.