Neural image processing strategies applied in real-time pattern recognition

Samples from stochastic signals with sufficient complexity need to agree in only a small portion of their details in order to justify rejection of the hypothesis that they arise from independent sources. The failure of a statistical test of independence can thereby serve as a basis for recognizing complex patterns, provided that they possess enough degrees-of-freedom and a representation can be found which extracts them, This paper describes a pattern recognition application of this statistical principle when conjoined with an image coding strategy that appears to be the representation used in the visual cortex of the mammalian brain. The result is a practical system for automatic real-time personal identification, based upon imaging the random texture visible at some distance in the iris of a person's eye. The recognition algorithm demodulates the iris texture with complex-valued 2D Gabor wavelets, and coarsely quantizes the resulting phasors to build a 256-byte 'IrisCode' whose entropy is roughly 173 bits, Ergodicity and commensurability facilitate extremely rapid comparisons of entire IrisCodes using 32-bit parallel integer-XOR instructions, Recognition decisions are achieved with astronomic confidence levels, and with exhaustive database searches conducted at the rate of about 10,000 persons/s. (C) 1997 Academic Press Limited.