Recognition of rotation, size and shape of gray-scaled human face by a rotation and size spreading associative neural network

摘要

We proposed a rotation and size spreading associative neural network. The characteristics of this neural net is to spread the information of axis orientation and size of the object on surrounding space by spreading weight that has similar tuning characteristics to axis orientation selective neurons and size recognition neurons in the parietal cortex. This neural net can recognize the orientation of the object irrespective of its size and shape its size irrespective of the orientation and shape, and its shape irrespective of the orientation and size. We expanded the input object from the binary image to the gray-scaled one. The rotation and size spreading associative neural network was able to recognize rotation, size and shape of gray-scaled human faces. Also, this neural net was able to recognize ambiguous figures (disoriented faces) and explain the decrease of recognition rate by face rotation.