Effective Feature Extraction Algorithm for Iris Recognition
Fei He · Journal of Jilin University · 2009
In order to avoid the shortcoming of non-dynamic base vectors and specific parameters in existing feature extraction based on linear transformation,after analyzing iris geometry features and recognition principles,an iris feature extraction using independent component analysis method is proposed.The algorithm eliminates the iris feature space redundancy furthest,and overcomes the flaw of non-dynamic feature base vectors in traditional linear transformation.The iris classification with BP neural network achieves lower dimension and effective feature expression.Experiments in small sample space using self-made iris database JLU-IRIS indicate the accuracy and validity of the algorithm through three groups of different recognitions rate 100%,96.5%,92.5%.