Abstract:
Matched filtering is an optimum detection process when the only difference between the transmitted and received signals is an additive Gaussian noise component. In practice, however, multipath and time dispersions in either deep or shallow water cause energy spreading that distorts the transmitted signal of an active sonar and this results in a loss of matched-filter processing gain [J.-P. Hermand and W. I. Roderick, IEEE J. Ocean. Eng. 18, 447--465 (1993)]. Standard ``sonar matched filtering'' ignores the propagation medium effects and is therefore suboptimum. This work investigates how the environmental and geometry parameters of detection problems affect the performance of the standard sonar matched-filter detector. Also, a replica-model-based matched filter is implemented that accounts for the time dispersive properties of the underwater environment by convolving the transmitted active signal with the transfer function of the medium. The proposed replica-model-based matched filter is an extension of matched-field processing techniques to active, time-domain problems and is an optimum detection processor if the environment and geometry are modeled accurately. The two methods, standard and replica-model-based, are evaluated with the use of receiver operating characteristic (ROC) curves. [Work supported by DREA and ONR.]