Embedded system for iris recognition using Harr wavelet transform

This paper presents the development of an embedded system for access control from iris recognition, implemented in Analog Devices ADSP-BF533 digital signal processor. The algorithm provides the ability to register users or validate the identity of external individuals using a technique of identification based on 2-D Harr wavelet transform, circular averaging filters, convex hull, geometric transformations, codification, quantification, and euclidean and Hamming distance. To evaluate the performance of the system we use 20 users, with 20 iris images each, taken from the CASIA-Iris database. We obtained a false acceptance rate of 0.25% and 5.5% false rejection, identifying the high efficiency offered by the calculation of the Hamming distance from the coefficients of the Haar wavelet transform, which added to the euclidean distance technique, offer high security by significantly reducing the possibility of access by unauthorized users.