A Neural Network Model for Chinese Character Perception

As we know, there is a debate on the mode of visual perception for a long time, that is, does the visual system acquire percept from local to global or from global to local. A three-layer perceptual model with function of signal decomposition and integration is developed, and Chinese character grapheme is chosen to train the model. The model is trained under different initial conditions and simulation results show that both of the perceptual modes may exist with different probability which is determined by the differentiation state of the neurons, perceptual object and the environment. Under the framework of complexity theory, further analysis reveals that the emergence phenomena exist in both of the perceptual modes.