A parallel analog signal processing unit based on radial basis function networks

This paper reports the design of an analog neuron circuit with Gaussian activation functions. The circuit is intended to be used for parallel computation of spatial distributed signals like the ones from array sensors (e.g. detectors) and images. The importance of the model is discussed in the context of Radial Basis Function and Regularization theory. Analog hardware implementation is performed as a natural approach for massive parallel computation of radial basis functions. Gaussian unit equations and simulation plots are provided along with an analysis of adjustable parameters.