Researchers introduce distinct hand database of individuals, which is acquired using a tailored hardware setup. The database contains four biometric traits: dorsal vein, wrist vein, palm vein, and palm print of the same person, which enables the multimodal biometric authentication exploration to create a spoof-proof authorization system. All 4 biometrics are captured using a single hardware device.
Abstract
“Biometric authentication is one of the most exciting areas in the era of security. Biometric authentication ideally refers to the process of identifying or verifying the user through physiological and behavioral measurements using security processes. Multimodal biometrics are preferred over unimodal biometrics due to the defensive nature of multimodal biometrics. This research introduces a distinct hand database of individuals, which is acquired using a tailored hardware setup. The database contains four biometric traits: dorsal vein, wrist vein, palm vein, and palm print of the same person, which enables the multimodal biometric authentication exploration to create a spoof-proof authorization system. All four biometrics are captured using a single hardware device. The veins are highlighted by lighting up the infrared (IR) light emitting diodes (LEDs) and a complementary metal oxide semiconductor (CMOS) image sensor is used to capture the vein image. Similarly, the palm region is equally illuminated with the array of white LEDs, and the palm print is captured using the CMOS image sensor. The CMOS sensor has a simple structure and uses a single camera. A total of 308 participants were included in the database, resulting in 8336 unique hand vein and palm print images of both hands. Compared the captured database with the existing databases such as PUT, FYO, VERA, and Bosphorus in terms of availability of the traits, number of subjects, total number of images, number of sessions, and gender. Preliminary experiments were conducted on the self-acquired database using an open-source software package named “Orange”. Various performance parameters were measured to construct a cost-effective authentication system. The classification accuracy obtained in k nearest neighbor (kNN), random forest (RF), support vector machine (SVM), neural network (NN), gradient boosting (GB), and logistic regression (LR) algorithms at a 70% learning rate are 99.8%, 99%, 99.8%, 99.8%, 97.7%, and 99.8%, respectively.”
Find the open access technical paper here.
Bharath M. R.and K. A. Radhakrishna Rao
Department of Electronics and Communication Engineering, PES College of Engineering, Mandya, Karnataka, 571401, India.
Visit Semiconductor Engineering’s Technical Paper library here and discover many more chip industry academic papers.
Leave a Reply