Research On Biological Lateral Inhibitory Mechanism And Its Application

Lateral inhibition is one of the basic information processing principles in the neural systems. Since the finding of lateral inhibition in the eye of Limulus, lateral inhibition has been discovered in many neural circuits, such as those in vertebrate somatosensory cortex, auditory cortex, visual system, hippocampus, and basal ganglia. Lateral inhibition can be used in many fields, such as image processing, pattern recognization, artificial intelligence, imaging navigation. In the paper, we don't only research on the theory of lateral inhibitiory mechanism, but also pay more attention on the application of lateral inhibitiory mechanism.Firstly, lateral inhibitiory mechanism is used in image processing. The common digital non-recurrent neural network with lateral inhibition and double-peak Gauss distributed non-recurrent neural network with lateral inhibition are applied to the image enhancement; the shunting recurrent neural network with lateral inhibition for edge detection is improved; besides, we combine plus couple neural network with lateral inhibitory mechanism for image segmentation. The processing results show that lateral inhibitiory mechanism improves the edge enhancement and connectivity of image.Secondly, a new method based on spatial-temporal information for the task of moving objects detection has been introduced. The algorithmic lateral inhibition model (ALI) is proposed to extract the motion in multiple channels in the form of temporal and spatial-temporal filtering. Then the noise is diluted by the spatial-temporal ALI. The experimental results show that the approach can obtain the silhouettes of objects accurately.Thirdly, we build a system of neural networks ensembles based on lateral inhibitiory mechanism of cortical columns, to simulate the neural firing phenomenon in cortical columns and be used for position tracking. Neural networks ensembles are composed of many Stein neurons or improved Hodgkin-Huxley neurons. Lateral inhibitiory mechanism of cortical columns is applied to the coding and decoding of neural networks ensembles to realize the negatively correlated firing. Neural networks ensembles based on lateral inhibitiory mechanism of cortical columns improve the accuracy of position tracking and the stability of systems greatly.Finally, recurrent neural network with lateral inhibition is proved by theory and experiments to play the role of winner-take-all. Then the function of winner-take-all is also carried out by new neural networks ensembles based on lateral inhibition of cortical columns.