"Deep Hide" Target Detection/Recognition Using Pseu

An innovative approach is proposed for the analysis of nonlinear and/or nonstationary signals and systems/sources that generate them. This approach is based on a canonical functional representation which yield multidimensional measurements of the nonlinear and/or nonstationary characteristics and allow system identification beyond usual linear and time-invariant (stationary) models. This approach can be applied to acoustic, electromagnetic and communication signals to extend recognition/cIassification capabilities beyond traditional linear time/frequency domains and reduce false alarm rates. The proposed techniques\' utility/povver will be demonstrated by deteiled analytical derivations, computer simulations, and test results on acoustic end radar data representing detection/classification tasks of interest to the Navy. MultiSpec has considerable experience with nonlinear and/or nonstationary analysis of real data in areas of importance for the Navy. UItra-wideband radar is enhanced when converted into a system identification task and higher-order signal analysis tools are used for target characterization. Data from ultra-vvideband radar discrimination experiments using targets of metal, clutter, and radar absorbent materials are shown in terms of bispectra and nonlinear kernels. Theory, simulations, and actual results from underwater acoustic data shovv the achievable with our new techniques for nonstationary analysis of transient signals.