Improving the Performance of Particle Swarm Optimization for Iris Recognition System Using Independent Component Analysis

Recently, iris recognition had been gained growing interest from researchers because of its high accuracy against other person identification techniques. This work presents an iris recognition system based on particle swarm optimization (PSO) and Independent Component Analysis (ICA) as feature extraction algorithm. Many experiments were conducted using different: swarm size (20, 40 and 80), PSO iterations (100 and 200) and ICA feature vector length (64, 128, 256 and 512). The results showed that: the best performance of the iris recognition system (high PSNR, high recognition rate and lower MSE) will be increased when increasing the ICA feature vector length to 512, increasing the swarm size to 80 and decreasing the number of iterations to 100. Best obtained value of recognition rate is 98%, best PSNR value is 38 and lower MSE value is 0.0011. The suggested iris recognition system has the ability to recognize un trained iris images but with lower performance.