7 June 2011 Spotting and tracking good biometrics with the human visual system
Author Affiliations +
Abstract
We mathematically model the mammalian Visual System's (VS) capability of spotting objects. How can a hawk see a tiny running rabbit from miles above ground? How could that rabbit see the approaching hawk? This predatorprey interaction draws parallels with spotting a familiar person in a crowd. We assume that mammal eyes use peripheral vision to perceive unexpected changes from our memory, and then use our central vision (fovea) to pay attention. The difference between an image and our memory of that image is usually small, mathematically known as a 'sparse representation'. The VS communicates with the brain using a finite reservoir of neurotransmittents, which produces an on-center and thus off-surround Hubel/Wiesel Mexican hat receptive field. This is the basis of our model. This change detection mechanism could drive our attention, allowing us to hit a curveball. If we are about to hit a baseball, what information extracted by our HVS tells us where to swing? Physical human features such as faces, irises, and fingerprints have been successfully used for identification (Biometrics) for decades, recently including voice and walking style for identification from further away. Biologically, humans must use a change detection strategy to achieve an ordered sparseness and use a sigmoid threshold for noisy measurements in our Hetero-Associative Memory [HAM] classifier for fault tolerant recall. Human biometrics is dynamic, and therefore involves more than just the surface, requiring a 3 dimensional measurement (i.e. Daugman/Gabor iris features). Such a measurement can be achieved using the partial coherence of a laser's reflection from a 3-D biometric surface, creating more degrees of freedom (d.o.f.) to meet the Army's challenge of distant Biometrics. Thus, one might be able to increase the standoff loss of less distinguished degrees of freedom (DOF).
© (2011) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Harold Szu, Harold Szu, Jeffrey Jenkins, Jeffrey Jenkins, Charles Hsu, Charles Hsu, } "Spotting and tracking good biometrics with the human visual system", Proc. SPIE 8058, Independent Component Analyses, Wavelets, Neural Networks, Biosystems, and Nanoengineering IX, 80580T (7 June 2011); doi: 10.1117/12.887520; https://doi.org/10.1117/12.887520
PROCEEDINGS
12 PAGES


SHARE
Back to Top