TRANSIENT PROCESSING USING THE GABOR REPRESENTATION

CONVENTIONAL SONAR PROCESSING IS BASED ON TECHNIQUES DEVELOPED FOR NARROWBAND STATIONARY (LONG DURATION) SIGNALS. THE INADEQUACIES OF THESE TECHNIQUES WHEN APPLIED TO THE DETECTION AND CLASSIFICATION OF TRANSIENTS MOTIVATES THE SEARCH FOR NOVEL APPROACHES TO THE PROBLEM. THE GABOR REPRESENTATION IS A PROMISING TECHNIQUE FOR DETECTION AND CLASSIFICATION OF UNDERWATER ACOUSTIC TRANSIENTS, WHICH PROVIDES AN ALTERNATIVE TO "TRADITIONAL" FFT BASED PROCESSING. BASED ON PRELIMINARY RESULTS OBTAINED ON SYNTHETIC DATA AND USING OUR EXTENSIVE EXPERIENCE IN TRANSIENT PROCESSING, WE BELIEVE THAT THE GABOR REPRESENTATION IS A NATURAL WAY OF CHARACTERISING UNDERWATER TRANSIENTS. IN PHASE I WE WILL PROCESS REAL ACOUSTIC TRANSIENTS USING BOTH CONVENTIONAL AND GABOR BASED TECHNIQUES, AND PERFORM A COMPARATIVE EVALUATION. WE WILL ALSO COMPLETE THE DEVELOPMENT OF TRANSIENT DETECTION AND CLASSIFICATION ALGORITHMS USING THE "GABORGRAM" AS A STARTING POINT. IN PHASE II WE WILL DESIGN AND BUILD A LOW-COST REAL-TIME TRANSIENT PROCESSING WORKSTATION TO BE USED FOR TEST AND EVALUATION OF GABOR PROCESSING ON LARGE VOLUMES OF DATA, FOR R&D, AND AS AN ADJUNCT TO OPERATIONAL PROCESSING SYSTEMS.