Receiver operating characteristic curves for linear arrays of vector sensors using nonlinear cardynull processing

Abstract

During the past decade researchers have been considering vector sensors for use in linear towed arrays for passive target detection. Linear processing is often used due to its simplicity and significant directivity improvements. Nonlinear processing holds the potential for further directivity improvements; however, it also presents the risk of amplifying uncorrelated noise. This thesis simulated a correlated signal in uncorrelated noise to investigate the potential of a nonlinear (but non-adaptive) processing technique. It demonstrates that the increased directivity and substantially diminished response from the ambiguous direction is quite beneficial when the signal is located within certain quadrants. It also demonstrates that linear processing is more effective than this nonlinear processor near endfire. In all cases, the signal to noise ratio was high enough to be detectable by basic array gain from multiple sensors. Monte Carlo simulations were completed to generate detection statistics and ROC curves were created to illustrate the relative effectiveness of: pressure-only sensor arrays, linearly processed vector sensor arrays, and nonlinearly processed vector sensor arrays. For a broadside signal in uncorrelated noise, simulations indicate an array with eleven vector sensors can achieve a 3 dB improvement if the nonlinear processing defined in this thesis is utilized instead of linear processing.