Offline Gurmukhi script recognition using knowledge based approach & Multi-Layered Perceptron neural network

The research area of handwriting recognition attracted number of researchers because of the challenges exist in it. The main difficulties in this area of research are: variations in handwriting styles of different writers, complex behavior of different languages used for writing etc. This paper focuses on Offline handwriting recognition process for an Indian language “Punjabi”. The script used to write Punjabi in India is Gurmukhi script. This paper presents a system which recognizes the handwritten words of Gurmukhi script. The technique used for recognition purpose in this work is based on Multi-Layered Perceptron (MLP) neural network. This paper presents a three layered architecture of neural network consisting Input layer, Hidden layer or training layer and Output layer. The segmentation phase in this work is implemented, using knowledge based approach. In segmentation phase, the success rate of 94.87 % is achieved to extract characters from Gurmukhi words and the overall recognition rate is observed as 82.06% for complete Gurmukhi words.