Hand-written Malayalam character recognition an approach based on pen movement

In this paper we introduce a novel approach for character recognition based on the pen movement i.e., recognition based on sequence of pen strokes.A Backpropagation Neural Network is used for identifying individual strokes.The recognizer has a two-pass architecture i.e., the inputs are propagated twice through the network.The first pass does the initial classification and the second for exact recognition. The two-pass structure of the recognizer helped in achieving accuracy of about 95 percent in recognizing Malayalam letters.The training set contains samples of all independent strokes that are commonly used while writing Malayalam.Input values to the network are the directions of pen movement.A “minimum error” technique is used for finding the firing neuron in the output layer. Based on the output of FirstPass the network is dynamically loaded with a fresh set of weights for exact stroke recognition.Analyzing the stroke sequences identifies individual characters.This work also demonstrates how a statistical pre-analysis of training set reduces training time