Personal Identification Using Combined Biometrics Techniques

Demand for reliable biometric systems has highly increased during that past decade. However, despite of the efforts conducted in the biometrics field, there is still a possibility of successful fraud attempts. Institutes and large organizations in attempt to improve the systems false acceptance rate (FAR), depend on the concept of using more than one biometric feature to positively identify a person. This technique is referred to as combined biometric. This article presents a design of a low cost biometric system that uses combined biometric features to improve the overall system performance by decreasing the false acceptance rates (FAR) and false rejection rates (FRR). The system creates a unique template for each subject based on both his/her facial expressions and ear patterns. Discrete cosine transformations are employed as a feature extraction method for the system then the template is fed into a back propagated feedforward neural network to classify and identify the target. We demonstrate experimentally, that by using only 4 input images as training images, fed into a two layer neural network with 20 input neurons an overall system success rate of 96.67% can be achieved.