Simulate Human Visual Perception Using Expert Neurons

This paper presents two main concepts. The first one is a new concept of artificial neuron, different from the classic approach. The second one is a proposal of a system that simulates the human visual perception in artificial intelligence systems based on what is known at this date on the human visual system at neurological and cellular levels. After studying the main two directions in the area, the neural networks (NN) and the expert systems (ES), a hybrid approach has been developed. This approach consists in the creation of a new type of neuron that was called expert neuron (XN). The main difference between the classical neuron and the expert neuron is that the expert neuron acts as an expert system at micro level and bases its activation on the decision made by an embedded inference engine and a dynamically defined knowledge base (KB). At macro level the system consists of several layers of neural networks containing only interconnected expert neurons. These layers follow as close as possible the structure of the human visual system, starting from the eye and ending with the visual areas from the cortex. An additional layer has been added which corresponds to a memory layer from the frontal cortex. The result was the CHILDREN system - Computer Human Interface for Learn Diagnose and Reasoning with Expert Neurons.