Echo inversion and target shape estimation by neuromorphic processing

Heuristic extension to make the neural net processor more neuromorphic by introducing nonlinearity is discussed and digital reconstructions with this extension are shown; these reflect noticeable improvement in image quality.

Abstract

A neural net processor is described for echo inversions and target shape estimations from incomplete frequency response data. The processor accomplishes the inversion and estimation by minimizing an energy function which bears information about the measured data, as well as the relationship between the target shape function (image) to be reconstructed and its frequency response. An iterative algorithm is developed for the processor to minimize its energy function to give the desired image as its neural state outputs. Successful digital reconstructions with the neural net processor using microwave radar imaging data are presented and an opto-electronic implementation of the processor is described. Heuristic extension to make the processor more neuromorphic by introducing nonlinearity is discussed and digital reconstructions with this extension are shown; these reflect noticeable improvement in image quality.