Tamil handwritten character recognition system enormously depends on its character features. This paper deals with the feature extraction and the three ways of feature predictions that are experimented in order to grasp features from various Tamil characters possessing variations in style and shape. Shape, shape ordering and location-based instances are the features predicted from the characters. The key features of this paper are the strip tree-based hierarchical formation which deals with the shape features of the characters, the implementation of the Z-ordering algorithm for addressing the structure ordering and finally the representation of PM-Quad tree that deals with extracting locations of the character features. A hierarchical classification algorithm based on support vector machine is used for predicting the character from its character features using divide-and-conquer procedure. Proof of this work shows that this work can address more characters and its varied shapes.
展开▼
