Statistical Feature Extraction and Recognition of Isolated Handwritten Gujrati Characters

Indic Script Recognition is a challenging task due to the large character sets of different patterns and constraints posed by types of handwriting, the number of writers, the size of the vocabulary and the spatial layout. Recognition of Gujarati script is a less explored area and a very little attempt is made for recognition of only subset of handwritten Gujrati characters. This paper proposes a reasonable approach to recognize a subset of isolated Gujrati Handwritten characters. The character images to be recognized undergo preprocessing operations such as resizing, conversion to binary format, noise removal etc. Then statistical features such as Euler value, Standard deviation and Euclidian distance are computed. These statistical parameters are compared with the images of stored characters for feature matching. For recognition, an evaluation was done for each letter in the learning set and efficiency of recognition was evaluated for each letter. An average overall Recognition Efficiency of 56.14% is obtained. Recognition rate is highly affected by similarity of various characters. Analysis of characters where the recognition fails reveals that the visually similar symbols confuse the recognition system.