Abstract:

In view of the shortcomings of redundant data leading to computational complexity of feature extraction after traditional normalization in iris recognition and weakening of main texture information, we introduced the ROIbased iris normalization means and used PCA to extract principal component to reduce the dimension and remove noise further, then trained feature vectors using ICA to make ICA/PCA iris feature extraction algorithm; finally used cosine distance classifier to identify testing samples. Experimental recognition rate was reached up to 98.11%, and recognition time was less than 70 ms.

Key words: pattern recognition； iris recognition； region of interest(ROI)