Score Level based Fusion Method for Multimodal Biometric Recognition using Palmprint and Iris

Digital technology is always in our lives and hence security using biometrics has become inevitable in areas like banking, logical access control, law and enforcement etc. Person recognition using more than one biometric trait is known as multimodal biometric recognition. In this work, images of palmprint and iris are chosen as the traits for performing multimodal recognition. From these images the Features from these images were extracted using Log-Gabor transform, Histogram of Oriented Gradients (HOG) and Local Binary Pattern (LBP). Distance measures such as Hamming distance, Chi-Square distance and Euclidean distance were employed for generating matching scores. A weighted sum score level fusion technique was applied for combining the scores generated from iris and palmprint images. The performance of the system was evaluated by plotting the Detection Error Trade off (DET) curves and Receiver Operating Characteristic (ROC) curves. The proposed system proved that multimodal recognition performed best than unimodal with a recognition accuracy of 92.23% by employing HOG as the feature descriptor. It was also found that less error rates to the system was achieved with Log Gabor filter.