Multi-biometric System for Security Institutions using Wavelet Decomposition and Neural Network

Abstract

Biometric systems are currently considered one of the leading methods for security and access control systems. The use of multi-biometric in verification and identification provides more reliability and accuracy for such systems. In this paper three biometric traits have been used face, iris and fingerprint for identification purpose. After preprocessing feature extraction for each trait, wavelet decomposition was used. Back-propagation neural network was employed for the training of the system. The results showed a highly accurate recognition rate after 298 epoch of training. A measurement of MSE and PSNR. With false acceptance rate (FAR) of 0% and false rejection rate (FRR) of 3% were calculated for system performance evaluation.