Low Volume, Low Power, Real Time Image Processing

"Irvine Sensors Corporation's (ISC) massively parallel Three-Dimensional Analog Neural Network processor (3DANN) brings near human-level data processing capabilities for applications in open-ended problems like pattern recognition, clutter discrimination,and target tracking and identification. The approach emulates the massive parallelism inherent in the brain to achieve extremely high computational performance at very low power and small volume. The core 3DANN processor is a general-purpose analogconvolution engine that can perform over 1 TeraOp inner product calculations per second. It occupies less than 1.5 cm3 volume and consumes < 10W of power. Since convolutions form the basis of most image processing and pattern recognition applications,3DANN holds orders of magnitude advantage over conventional digital processors in terms of computational throughput, size, weight, and power consumption. 3DANN based systems bear potential to become PetaOp speed pattern/data-fusion processors that canrapidly process multi-megabit data streams from multiple sensors, perform rapid analysis, and tirelessly provide a course of action in environments that are hostile to human operators. The proposed innovation will find application in unmanned robotic ATRsystems and embedded biometric scanners and face recognition systems."