Skip to Main Content
Creating a biometric verification system in a resource-constrained embedded environment is a challenging problem. This paper describes an efficient fingerprint verification module, which is part of an embedded device called ThumbPod. The whole fingerprint verification algorithm runs on a 50 MHz fixed-point processor. As the result of our SW/HW optimizations, we achieve 55.6% and 60.0% execution time reduction for the minutiae extraction and the matching, respectively, compared to a traditional implementation reference. The complete process finishes in less than 5 seconds.