Human face recognition by an association cortex - Entorhinal cortex-Hippocampal formation model

摘要

In regard to memorization and recollection process of visual stimuli, we proposed a neural network model, i.e. Association Cortex - Entorhinal Cortex - Hippocampal Formation Model (AEH model) and tested the recognition performance of AEH model using human faces. This model is constructed by incorporating neorophysiological knowledge into Jones's hypothesis. We introduced new neural net which performed a sparse code learning between long term memory layer in inferotemporal cortex and hippocampal formation in order to recollect the sparse code in hippocampal formation and to perform additional learning in recollection. We presented each human face to the AEH model twice and regarded that the AEH model recognized each face correctly if the AEH model learned the face the first time and recollected it the second time. We used 20 human faces to learn and recollect successively in our experiment. Recognition performance of the model was over 95 percent correct. Besides the model performed additional learning and reinforced memory in recollection.