Analog 3-D Neuroprocessor for Fast Frame Focal Plane Image Processing

A particularly challenging neural network application requiring high-speed and intensive image processing capability is target acquisition and discrimination. It requires spatio-temporal recognition of point and resolved targets at high speeds. A reconfigur able neural architecture may discriminate targets from clutter or classify targets once resolved. By mating a 64 x 64 pixel array infrared (IR) image sensor to a 3-D stack (cube) of 64 neural-net ICs along respective edges, every pixel would directly input to a neural network, thereby processing the infor-mation with full parallelism. Being mated to the infrared sensor array, the cube would operate at 90°K temperature with <250 nanosecond signal processing speed and a low power consumption of only -2 watts. For low power and compactness in hardware, the emphasis has been on parallelism and analog signal processing. A versatile reconfigurable circuit is presented that offers a variety of neural architectures: multilayer perceptron, template matching wit...