Two-D position invariant recognition of object orientation and shape by a rotation spreading associative neural network

摘要

In the brain, parieto-hippocampal spatial recognition (and memory) system work in parallel with the object recognition system. It is well known that neurons in the parietal cortex is related to the recognition of 3-D object shape as well as to the recognition of spatial factors. Paying attention to these findings, we proposed a rotation spreading associative neural network (RSAN net) which spread the spatial information of an object by a spreading neural layer. This neural network has the shape invariant orientation recognition and the orientation invariant shape recognition characteristics of human faces. We examined the recognition characteristics of the RSAN net in two-D space. We then developed a position invariant RSAN net that simultaneously recognized the rotation and shape in two-D space. The position invariant RSAN net was able to recognize the rotation (within the range from -60 to 60°), and shape (in all angles) in two-D space.