A New Approach for Sentence Recognition Using Multiple Features

This paper presents the work towards recognition of sentence by using segmentation based on bounding box for character recognition. After segmentation various variant and invariant features are extracted like Correlation method, Shadow feature, horizontal and vertical profiling, Chain code and Projection methods which increase the recognition capability of a character. Classification techniques are used for improving the recognition of sentence. Dictionary search for each sentence is performed, after first character is recognized and matched. The recognized sentence is spelled out by the system.