Abstract
A passive sonar algorithm measures the pressure field at a sensor array then estimates the sound source location from these observations and an acoustic propagation model. Passive sonar performance is usually characterized by the mean squared error (MSE) between the estimated and true source locations. Consequently, passive sonar performance bounds typically provide lower bounds on the achievable MSE for a given array and environment. Such MSE bounds can be misleading in environments with strong sidelobes, as the optimal estimator may choose an unlikely location to balance the error among several highly likely locations. An alternative algorithm would be to partition the search space, then use the array observations to choose which block contains the source, but not its exact location in the block. The resulting binary performance metric is now the probability of choosing the incorrect block or error probability (Pe). Information theory allows us to formulate Bayesian bounds on the minimum achievable Pe for a given array, propagation environment, and search space partition. These bounds quantify the trade-off among SNR, Pe, and location estimation accuracy for passive sonar. [Work supported by ONR Code 321US.]