Recognition of face position and shape by spread pattern of spatial spreading associative neural network
Keisuke Nakamura, N. Komatsu, N. Kinoshita, S. Miyamoto · 2002
Human beings can recognize the position and shape of an object (e.g., face) at once by the parallel information processing systems (i.e., spatial and shape recognition systems) in the brain. In the study, a spatial spreading associative neural network (SSAN net) that imitates the spatial recognition system in the brain is developed. The SSAN net recognizes the face position, shape and orientation. The SSAN net spreads the spatial information on the surrounding space by the spatial spreading neural layer to maintain the spatial discrimination of the input pattern. The generalized inverse learning and the population vector method are used for the recognition of the face position and orientation. The SSAN net recognizes the face shape irrespective of its position, and the face position irrespective of its shape. The SSAN net also recognizes the rotated face position and orientation by cell assembly of the face position and face view orientation neurons. The spatial spreading neural layer functions critically for the invariant position and shape recognition of the face.