Feature Level Fusion of Palmprint and Iris Images for Person Identification

Biometrics plays an important role in recent security applications. Unimodal biometric systems can?t perform well in certain applications due to the presence of noise in data collected by sensors. Moreover, inter-class similarities and intra-class variations may lead to misclassification. Multimodal biometric systems can be utilized to obtain promising identification ratios according to their higher accuracy. Multimodal systems not only improve performance but also reduce the problem of universality. In the present work, a feature level fusion technique that combines palmprint and iris images will be presented. The valley detection technique is used to extract the region of interest (RoI) of a palm. The neighbor-pixels value approach (NPVA) is used to extract the RoI of an iris. Discrete wavelets transform (DWT), Fast Fourier Transform (FFT), and discrete cosines transform (DCT) are used to extract discriminating features. Connectivity points and lifelines orientations are used to extract more palmprint features, whereas wavelet transform is used to obtain more iris features. A neural network classifier will be used for classification purpose in addition to a minimum distance classifier. The accuracy of the proposed system was 99.85 and % which exceeds the accuracy of 97.09 while using unimodal palmprint classification and that of iris which reached 96.83%.