Handwritten Armenian character recognition based on discrete cosine transform and artificial immune system

Artificial Immune System[1] is engineering system which has been inspired from the functioning of the biologic al immune system. In this paper, handwritten Armenian character recognition strategy using artificial immune system was proposed and carefully experimented. With 90 feature coefficients extracted from 24*24 Armenian character image using DCT based on 8*8 image sub-block as its feature vector, 38 antibody libraries for 38 character category were trained and built to recognize Armenian characters with artificial immune algorithm. The contrast experiment was done using three-tiered feed-forward, back-propagation neural network model with sigmoid transfer function, 0.01 learning rate parameter and the same input feature coefficients[8]. The experimental results indicated that the artificial immune system model has more advantages than BP neural network in character recognition.