Correlation Computations for Movement Detection in Neural Networks

The visual information is inputted first in the retina of the biological network. Reichard[1] described that the autocorrelation is a principle for the evaluation of sensory information in the central neural system. Retinal ganglion cells produce two types of responses: linear and nonlinear. The nonlinear responses are generated by a separate and independent nonlinear pathway. The nonlinear pathway is composed of a sandwich model in the neural filters. It is important and useful to clarify the structure and the function of the network with linear pathway and the nonlinear pathway. In this paper, we show the auto and cross correlations play the important role in the sensory movement stimulus by analyzing the neural network with the linear and the nonlinear pathways.