Bayesian passive sonar tracking with conventional beamformer-level data

A likelihood function for a multi-sensor, passive sonar Bayesian tracker may use data directly from the array elements or, alternatively, from the output of a conventional beamformer (CBF). Here, we compare the performance of a Bayesian tracker when using element- versus CBF-level data. We observed that, provided the CBF's beams are sufficiently closely spaced, the tracker's performance with CBF-level data is similar to when it is using element-level data. However, when the spacing between CBF beams becomes too large, tracker performance with CBF-level data degrades significantly, particularly when the target is located between the maximum response axes of two adjacent beams.