Stable reciprocal image associations in cognitive systems

Sensory inputs such as visual images or audio spectrograms can act as symbols in a new cognitive model. The stability of direct image association operators allows the discrete bit patterns in a general-purpose symbol processing system to be replaced with continuous real-world signals. Analogous to an SR flip-flop, two reciprocal images recursively connected by association processors, can ldquolockrdquo each other in place. A computational model of the Brodmann areas, whose boundaries are defined by the thickness of the exterior and interior lamina in cerebral cortex, closely resembles this structure. The recurrence between the cells in the cortical columns allows local connections in small regions to form overall global image associations. An implementation, based on neurotransmitter field theory, demonstrates the stability of the reciprocal-image attractors.