Bounds on bearing estimation precision for binaural broadband active sonar

A broadband active sonar transmits sound pulses and then listens for the received echoes from these pulses that are being reflected by the target. The detection of an echo signal coming from a possible target is done by using matched filtering the echo with the transmitted signal and once the target is detected, the next phase is target localization. In target localization phase the active sonar estimates the target’s location in terms of range and bearing where range is estimated from time delay between transmission of pulse and reception of its echo. This thesis focuses on the bearing estimation part of the broadband active sonar. The conventional way to estimate target bearing is to use a receiver array from the bearing information embedded in the differential time delay of received signal at each sensor. However, several recent works focused on the analysis of received echo where the bearing information is encoded by the angle-dependent beampattern of the transmitter. All the previous related works used mainly monaural receiver system. This thesis focuses on the binaural model for a broadband active sonar which analyzes the bearing Fisher Information (FI) and determines the performance bound of bearing estimation by using Cramer-Rao Lower Bound (CRLB). The CRLB on target bearing has been evaluated as a function of SNR, angle and interaural distance which provides an insight for how FI varies with angle for different interaural distance. Monte Carlo trials were also performed to simulate the Mean Squared Error (MSE) from a Maximum Likelihood Estimator (MLE) to validate the CRLB on target bearing. The CRLB of target bearing is consistently reduced with increasing interaural distance which indicates performance improvement. The monaural active models suggested that there is a sweet spot where the FI is maximized and bearing estimation performance is optimum. However, for increased interaural distance which provides best estimation performance, this binaural model suggests that the FI increases monotonically to its maximum as the target moves towards broadside. In this work by introducing a binaural active sonar model, the bearing information available from both the angle-dependent transmitter beam pattern and the relative phase difference between the received signals on the two sensors have been utilized to measure the performance of this model.