A view-based approach for recognition of Bengali printed characters

This paper presents a method for recognizing Bengali printed characters by using view-based and layer-based approaches. Two different view-based approaches, the top-bottom and the left-right have been used. The layer-based approach is also considered here. No thinning or segmentation is required. The individual character is taken as a whole image. The characteristic points are extracted from the views after some preprocessing. These points are then used to form a feature vector which represents the given character. These feature vectors are classified with the aid of dynamic time warping distance, k-NN and ensemble of classifiers. The concept can also be extended for Bengali word recognition.