Compact Neural Network Architecture and Algorithms for Signal Processing

In order to improve the performance and capability of missile defense systems and missile guidance systems, Celnet Technology Corporation proposes to develop a novel neural network computer architecture and algorithms for signal processing and decoding in demanding communication applications. The proposed locally connected neural network utilizes hardware annealing to solve optimization problems in real time signal processing. Key features of the proposed system include compact VLSI implementations, low power dissipation for portability, high speed analog continuous time operation, while still achieving robust decoding in very noisy environments. Missile defense systems in the digital battlefield involving complex communication patterns would benefit from the increased performance and reliability in the communication links. In Phase I Celnet will prove the feasibility of this approach, and will provide designs and architecture for the system integration.