Adaptive sub-band processing in active detection and tracking

Abstract

Broadband active systems have the potential to increase detection performance in reverberation-limited areas through increased signal bandwidth, i.e. small resolution cell. However, large bandwidth could over-resolve the target and segment the echo energy, as well as raise the false contacts, and therefore increase the false alarm rate. This effect becomes more apparent in shallow water channels that support multi-path propagation. One way to overcome this problem is to obtain a priori information regarding the channel characteristics and, based on that, utilize frequency agile methods to enhance detection performance for the particular environment. More specifically, in this paper we use multipath estimates to predict the most effective bandwidth to be utilized in an adaptive sub-band processing detection algorithm. Based on this prediction, a series of non-overlapping band-pass filters are applied to the full bandwidth complex matched filter output. The results are then combined at the contact level to reduce false alarms and enhance detection performance compared to the full bandwidth.

This methodology is applied to a real data set that includes a surface target and a synthetic submerged target. Comparison between sub-band and full-band processing

demonstrates performance enhancement by false alarm reduction while maintaining comparable probability of detection, therefore leading to improved Receiver Operating

Characteristic (ROC) curves. The full-bandwidth versus the sub-band processing is also evaluated at the automated tracker level.