‫التعرف على الوجه الثابت بإستخدام خوارزمية الإنتشار الخلفي بالشبكة العصبية‬

Abstract

Abstract The urgent need today in most business sectors (companies, organizations and institutions) systems to protect and control their employees (enter the employee to the labor sector, attendance, etc.) where he managed the physical characteristics of the human person, which distinguish it from among his own race some of the features and the so-called biometric (Biometrics). Where to humans, many of these biometrics, for example, including fingerprint, face, fingerprint, hand print, iris scan, fingerprint ear, footprint, fingerprint sweat. Allow systems to adapt to the classification and verification of these rights through biometrics. Researcher dealt with a system designed to distinguish people by recognizing the still face print; where they are including pictures of the faces of the people (known or unknown) and then classified, according to the researcher using the algorithm (BPNN) neural network with back Backpropagation Neural Network by train the classes of faces are then made to test the faces of the network is trained to classify these faces. The researcher evaluated the system by conducting tests on three databases record the Japanese Female Facial expression (JAFFE) Database that was captured in the Department of Psychology at the University of Kyushu, Japan, and ORL Created at the University of Cambridge in America and Aberdeen that have been created at the University of Stirling in Scotland, and was train the system to 302 image of the face (males and females) vary in the degree of lighting and direction of the face and facial expressions. and may the researcher tested the system on 451 image of the face have been identified on the 393 image and the percentage of classification correct for testing 87.1%, while the percentage of recognition on the faces are not training classes, including 6.25% (False acceptance rate), where the test samples 80 pictures. while the False rejection rate by 12.86% where deporting rate is (0.6), therefore, ensure that the identification process in the system allows the identification of which train samples of him, as the proportion of identified right on the face increases the more training classes.