Iris identification based on fourier coefficients and singular value decomposition

Abstract

Nowadays, both personal identification and classification are very important. In order to identify the person for some security applications, physical or behavior-based characteristics of individuals with high uniqueness might be analyzed. Biometric becomes the mostly used in personal identification purpose. There are many types of biometric information currently used. In this research, iris, one kind of personal characteristics is considered because of its uniqueness and collectable. Recently, the problem of various iris recognition systems is the limitation of space to store the data in a variety of environments. This research proposes the iris recognition system with small-size of feature vector causing a reduction in space complexity term. For this experiment, each iris is presented in terms of frequency domain, and based on neural network classification model. First, Fast Fourier Transform (FFT) is used to compute the Discrete Fourier Coefficients of iris data in frequency domain. Once the iris data was transformed into frequency-domain matrix, Singular Value Decomposition (SVD) is used to reduce a size of the complex matrix to single vector. All of these vectors would be input for neural networks for the classification step. With this approach, the merit of our technique is that size of feature vector is smaller than that of other techniques with the acceptable level of accuracy when compared with other existing techniques.