Six Biometrics Identification System (SBIS) Using Feed Forward Back Propagation Neural Network

Abstract

Generally, Biometrics identification system using for person identity depend on individual characteristics. Which, by its nature divided into two parts (physiological and behavioral). This system may be named as Automated Biometrics Identification System (ABIS) [18]. Designing ABIS as a search technology, this methodology is to implement a one-to-many query pattern convergence to patterns in a database containing many biometric templates. This operation named as biometric identification. So, it permits to compare a live-sample opposite the saved templates to find a match of a person and verify his identity. Unlike the other one-to-one identification system such as (one biometric template, one user sample and one authentication system) used in recognition system. In this paper, a new design was putted forward of combining six biometric identification system. (SBIS) is a biometrics system gathering six bio-trait physiological (face, iris, hand shape, thumbprint) and behavioral (signature and voice) respectively. These biometrics are collected from 18 individuals (16 for training 2 for testing) then arranged in separately dataset for each one, in a certain way, that guaranteed later when collecting the characteristics resulting from feature extraction operation after processing them separately in a big matrix that they belong to the same person [4]. Wavelet used as feature extraction filter, then assembling matrix feed to the backpropagation neural network with Levenberg-Marquardt as training function, the network trained with 50 and 100 hidden neurons respectively. The results are 95.3% and 98.44% in accuracy for 64 trained sample. And 100% with untrained sample. The aims of (SBIS) are improving security with high efficiency.