Active Sonar Target Classification in Littoral Waters

"Attempts to design target classifiers for active sonar have failed to identify reliable features for separating targets and clutter, causing performance in littoral waters to be severely degraded by clutter induced false alarms. The proximate cause offailure is the large number of possible return waveforms that correspond to a target at different possible aspects in complex, multi-path acoustic channels. Further, conventional matched filter techniques cannot distinguish between channel-induced andtarget-related features in the received signal. By introducing a new and unique replica set to properly decompose the return waveforms, we have solved these problems without recourse to matched-field processing. The resulting time-eigenvalue displaysprovide between 10 to 20 dB separation between targets and selected clutter. False alarms can be suppressed by several orders of magnitude for a given probability of detection. We have already achieved the Phase I objectives, including making the teststatistics robust against non-stationary contamination. The algorithms can run on existing and planned hardware, they are robust to environmental and channel changes, and require no operator intervention. We can now proceed to a demonstration of thetime-eigenvalue displays using mid-frequency, broadband waveforms. A 10 to 20 dB improvement in target-clutter separation with the ability to classify targets. The techniques translate directly to radar, c