Abstract
We have proposed in this paper an embedded palmprint recognition system using the dual-core OMAP 3530 platform. An improved algorithm based on palm code was proposed first. In this method, a Gabor wavelet is first convolved with the palmprint image to produce a response image, where local binary patterns are then applied to code the relation among the magnitude of wavelet response at the ccentral pixel with that of its neighbors. The method is fully tested using the public PolyU palmprint database. While palm code achieves only about 89% accuracy, over 96% accuracy is achieved by the proposed G-LBP approach. The proposed algorithm was then deployed to the DSP processor of OMAP 3530 and work together with the ARM processor for feature extraction. When complicated algorithms run on the DSP processor, the ARM processor can focus on image capture, user interface and peripheral control. Integrated with an image sensing module and central processing board, the designed device can achieve accurate and real time performance.
Highlights
Biometrics is a technique to identify a human being by his/her physical or behavioural characteristics, which, compared to smart card and passwords, are not easy to lose or forget
While complicated algorithms could run on the Digital Signal Processing (DSP) processor, the ARM processor can focus on image capture, user interface and peripheral control
We have implemented in this paper an embedded palmprint recognition system using the dual-core
Summary
Biometrics is a technique to identify a human being by his/her physical or behavioural characteristics, which, compared to smart card and passwords, are not easy to lose or forget. Most biometrics techniques try to identify a person using his body parts or behaviors such as face, fingerprint, Sensors 2012, 12 iris or gait. It was not until recently that researchers started to pay attention to hand-based biometrics like palmprint [1] and finger-knuckle-print [2] for personal authentication. The inner surface of the palm normally consists of principal lines and wrinkles. Researchers found that these non-genetically deterministic patterns are even different for identical twins and they are very useful for personal identification. As a relatively new biometric trait, Zhang [3] pioneered the work of palmprint recognition. A typical palmprint recognition system contains a scanner, preprocessing, feature extraction and matching. Different methods might use various features and differ in the design of matchers
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
