A concept acquisition method based on visual perception

In a physicalist theory of mind, a concept is a mental representation, which the brain uses to denote a class of things in the world. They can help us classify newly encountered objects on the basis of our past experiences. But how can we get the concepts? As visual information account for about 80% of total perceptual information, this paper proposes a concept acquisition method based on visual cognition. Firstly, we give a new definition of concept based on the classic views. Then, we imitate the process of visual cognition to get the approximate locations of the salient proto-objects. By using the improved cerebral cortex learning algorithm, the concepts of the objects can be acquired. However, the whole model also has top-down process. The concepts learned in turn affect the former saliency-based region selection method. We demonstrate that the suggested method can not only yield dynamic and phase-smooth concepts, but also recognize objects with good precise. Our findings will facilitate further studies on human thinking.