USING NEURAL NETWORKS TO RECOGNIZE AGILE EMITTERS: A FEASIBILITY ANALYSIS

POWERFUL INFORMATION PROCESSING DEVICES CALLED, NEURAL NETWORKS, ARE CURRENTLY BEING APPLIED TO PROBLEMS IN SPEECH RECOGNITION, COMPUTER VISION, MULTI-DIMENSIONAL PROBABILITY DENSITY FUNCTION ESTIMATION, COMBINING UNCERTAIN KNOWLEDGE, AND OTHER DIFFICULT INFORMATION PROCESSING PROBLEMS. THE RESEARCH PROPOSED HEREIN IS AIMED AT ASSESSING THE POTENTIAL OF NEURAL NETWORKS IN SOLVING THE MOST DIFFICULT PROBLEM FACING MODERN NAVAL ELECTRONIC SUPPORT MEASURES (ESM) SYSTEMS, NAMELY, RECOGNIZING EMITTERS WITH TIME VARYING BASIC PARAMETERS IN REAL TIME IN A DENSE RF ENVIRONMENT. NETROLOGIC PROPOSES A RESEARCH PROGRAM CONSISTING OF THREE SUCCESSIVE STAGES: 1) PROBLEM DEFINITION -SPECIFYING THE PROBLEM EMITTER CHARACTERISTICS AS WELL AS SENSOR AND ENVIRONMENTAL PROPERTIES, 2) NETWORK CONFIGURATION -DESIGNING THE NEURAL-NETWORK AND ASSOCIATED CONVENTIONAL PROCESSING TO PERFORM THE RECOGNITION TASK, 3) NETWORK IMPLEMENTATION -SPECIFYING THE COMPUTATIONAL CHARACTERISTICS TO ENABLE A COST-EFFECTIVE AND PRACTICAL IMPLEMENTATION.