Telugu handwritten character recognition using adaptive and static zoning methods

Character recognition is one of the fields of research in pattern recognition. Recognition of hand-written characters can be done either On-line or Offline. Not much substantial work has been published in the past on the development of hand-written character recognition (HWCR) systems for Telugu text. However none of them give 100% accuracy in recognition of Telugu characters. Therefore, it is an area of ongoing research. Our effort is intended to improve the accuracy in Telugu character recognition. This motivated us to undertake this work. Zonal based feature extraction is used in the present proposed work. We presented two methods for this purpose. First method is based on Genetic Algorithm and uses adaptive zoning topology with extracted geometric features. In second method, zoning is done in static way and uses distance, density based features. In both the contexts, we use K-Nearest Neighbor (KNN) algorithm for classification purpose. The character image is divided into predefined number of zones and features of all the zones in the image form a feature vector which is used in classification phase of hand-written character recognition. Using first method we obtained accuracies of 100 percent and 82.4 percent for 11 and 50 symbols respectively. Using second method we obtained accuracies of 100 percent and 88.8 percent for 11 and 50 symbols respectively.