ADAPTIVE PRE-PROCESSING AND TWO-DIMENSIONAL FILTERING FOR SONAR DISPLAYS

EACH PIXEL IN A CONVENTIONAL PASSIVE SONAR DISPLAY USES ONLY PART OF THE INFORMATION THAT IS AVAILABLE FOR DETECTION. THE ESTIMATORCORRELATOR DETECTOR COMBINES THE DIFFERENT PIXELS BY MEANS OF A WIENER FILTER, AND IT USES ALL AVAILABLE (OR HYPOTHESIZED) INFORMATION ABOUT THE DISTRIBUTION OF SIGNAL ENERGY IN FREQUENCY, TIME, AND BEARING. PERFORMANCE AND ROBUTNESS OF THE OPTIMUM DETECTOR CAN BE PREDICTED WITH THE HELP OF A SPECIALLY DESIGNED COMPUTER PROGRAM. IMPLEMENTATION IS GREATLY SIMPLIFIED AND PERFORMANCE IS OPTIMIZED WHEN AS MUCH ENERGY AS POSSIBLE IS CONCENTRATED IN THE SMALLEST POSSIBLE NUMBER OF STATISTICALLY INDEPENDENT PIXELS. A FEEDBACK MECHANISM ADJUSTS OUTPUT SAMPLING RATES AND BANDWIDTHS OF CONVENTIONAL FFT-BASED PRE-PROCESSING FILTERS TO OBTAIN THIS CONDITION. IDEALLY, AN ADAPTIVE KARHUNEN-LOEVE FILTERING ALGORITHM IS REQUIRED, AND IMPLEMENTATION OF SUCH FILTERS IS INVESTIGATED. THE WIENER FILTER BASED ESTIMATOR-CORRELATOR IS GENERALIZED TO NON-GAUSSIAN DATA, MULTIDIMENSIONAL DATA, AND VERY HIGH RESOLUTION DATA THAT ARE TOO DETAILED TO BE COMPLETELY DISPLAYED. THE PERFORMANCE OF A HUMAN OBSERVER IS EMULATED BY ADAPTIVELY ADJUSTING HYPOTHESIZED SIGNAL POWER AS A FUNCTION OF TIME AND BEARING.