Sonar target detection and classification against non-stationary interference using dynamic time-frequency localization

We address the well-known challenge of detecting a target in non-stationary clutter in active sonar using dynamic time-frequency localization. The challenge is to track a target against non-stationary reverberation from the bottom and sea surface, as well as backscatter from biologics. Definition of "target" is application-specific, e.g. in a Naval application the target would typically be a submarine or mine, in another application the target is an environmental AUV sensor tracing pollutant flow along the coast. Examples of non-targets include bathymetric features, wrecks, fish schools, other AUVs, etc. We present a concept for an adaptive subspace-tracking algorithm that navigates the time-frequency space of the target echo dynamically to provide enhanced target detection and classification. Simulation results are presented.