COMMUNICATIONS SIGNAL RECOGNITION AND DEMODULATION VIA NEURAL NETWORKS

THE GOAL OF THIS PROGRAM IS TO IDENTIFY SPECIFIC NEURAL NETWORK ARCHITECTURES AND ALGORITHMS WHICH ARE APPLICABLE TO COMMUNICATIONS PROBLEMS AND EVALUATE THOSE ARCHITECTURES AND ALGORITHMS VIA SIMULATION AND MODELING. WE DEFINE HEREIN A SYSTEM CONSISTING OF TWO DIFFERENT NEURAL NETWORK PARADIGMS WHICH ARE INTEGRATED TO PERFORM AUTOMATIC IDENTIICATION AND DEMODULATION OF DIGITAL COMMUNICATIONS SIGNAL DATA. THE COMPOSITE NEURAL STRUCTURE WILL MAKE USE OF (A) A SELF ORGANIZING CLUSTERER NETWORK WHICH WILL MONITOR THE FORMATION OF THE DEMODULATED SIGNAL CONSTELLATION AND (B) A MULTILAYERED PERCEPTRON FOR DEMODULATION PARAMETER EXTRACTION. THE TWO NETWORKS THAT WE HAVE CHOSEN TO USE ARE BASED ON WIDELY ACCEPTED PARADIGMS. FURTHEERMORE, OUR PHASE I APPROACH MAKES USE OF COMMERCIAL OFF-THESHELF SOFTWARE WHICH MAY BE DIRECTLY EMBEDDED IN OFF-THE-SHELF INTEGRATED CIRCUITS IN PHASE II.