Merging face and finger images for human identification

Abstract

Human identification performance reported so far using face or finger images under certain conditions is good practice, however, there is still a great need for better performance in biometrics for use in video surveillance. One possible way to achieve improved performance is to combine information from multiple sources. Besides, such systems alleviate some of the problems that are faced by single biometrics-based systems like restricted degrees of freedom, spoof attacks, and unacceptable error rates. We present a prototype bimodal biometric identification system by merging face and finger images. A novel approach is adopted to merge biometric (face and finger) traits of an individual to one image (containing features of both), named merged pattern. The integrated features are then extracted with an adaptive artificial neural network. The proposed algorithm is shown to exhibit robustness in achieving better classification results with both good generalization performance and a fast training/test time using variable public domain databases. Sources of variability include facial expression, gender, individual appearance, tilt, lighting conditions, and occluding objects (hair, spectacles, etc).