Adaptiveness, Error Resilience and Robustness Validation of the Multimodal Hand Biometric Recognition System

Abstract

The multimodal biometric system based on palm print (PP), finger geometry (FG) and dorsal palm vein (DPV) modality is proposed, specifically for high-security applications. This paper aims to prove that the proposed multimodal biometric system is an adaptive, error-resilient and robust system. A novel 'Optimum Weights Algorithm' makes the system adaptive and provides the best possible accuracy. An erroneous database of 100 users is collected to check an error resilience and robustness of the multimodal system. PP, FG and DPV feature extraction algorithms are used to extract feature vectors for all three modalities. Accuracy prediction is made by plotting the ROC curve for the multimodal system and estimating GARmin from that ROC curve. It is observed that the accuracies of the FG and DPV modalities remain unaffected for the erroneous database; however, there is a small decrease in the accuracy for PP modality. The values of accuracies obtained for the PP modality with both the degradations, namely, chalk dust and fine dust are 97.50% and 98.55% respectively, with a very low FAR level of 0.0001. For the erroneous database, the proposed multimodal system provides an accuracy of 99.80%.