Study On Dorsal Hand Veins Authentication Based on A Variety Of Factors Includes Translational And Rotational Effect

Abstract

Biometric is automated method which helps in recognizing a person based on newlinephysiological or behavioural traits. Behavioural characteristics are related to pattern of newlinethe behaviour of a person such as gestures, voice, signature etc while Physiological newlinecharacteristics are based on physical traits such as finger prints, face, DNA, hand and newlineretina, veins etc. Biometrics systems are preferred over traditional security methods like newlinepassword as these are subject to attacks which may lead to the loss of sensitive data. newlineHand vein identification system is drawing researchers concern as human veins are newlineunique, stable, permanent, and universal for every individual and are difficult to forge. newlineMoreover, the veins can be acquired through contact-less methods which has become newlinethe need of hour. newlineThe major challenge while developing the vein pattern based authentication system is newlinethe quality of image and versatility of vein pattern as these may cause performance newlinedegradation. So, here, a new method, Rider Cat-based Chicken swarm optimization newline(RCCSO) is proposed for hand vein recognition. Repeated Line Tracking method and newlineMaximum Curvature Points in Image Profiles are used for vein detection. Furthermore, newlinethe statistical features and CNN features are employed in Deep convolution neural newlinenetwork (DCNN) for classifying the hand vein. The training of DCNN is performed by newlineproposed RCCSO, which is designed by integrating Chicken swarm optimization and newlineCat swarm optimization in Rider optimization algorithm (ROA). The classification of newlinehand veins are based on a variety of factors, which includes translational and rotational newlineeffect in hand-postures along with thick veins due to age or disease. newline newlineThe proposed RCCSO-based DCNN provided qualitative performance with highest newlineaccuracy of 98.429%, highest sensitivity of 99.03% and highest specificity of 99.04%. newlineThe lowest errors obtained with the developed system are false acceptance rate (FAR) newlineas 0.97%, false rejection rate (FRR) as 0.96 % and equal error rate (EER) as 0.965% newline