Pattern Recognition in Alphabets of Oriya Language Using Kohonen Neural Network

Here a computerized reading of alphabets of Oriya language is attempted using the Kohonen neural network and its unsupervized competitive learning capacity as self-organizing map or the Kohonen feature map. The proposed pattern recognition does not treat a pattern as an n-dimensional feature vector or a point in n-dimensional space as is done in the traditional pattern recognition theory. We have tried with all the Oriya alphabets and have presented the study with respect to five of them in this paper along with their average distance per pattern in each cycle till we reach the permissible average distance. In the output picture the variation of the weight vector with respect to the alphabets is clearly observed.