Source Separation Using Time-Frequency and Range-Doppler Basis Functions with Sparse Estimation

A source separation scheme is proposed to estimate and isolate radar backscatter from communications interference. Building upon previous work in the area, the signal modeling spans an entire coherent processing interval while estimating scattering in delay-Doppler and communications interference in time-frequency. To maintain generality, the communications interference is modeled via a collection of time-frequency shifted Gabor functions of various resolutions. A Sparse Bayesian Learning routine is applied to the signal modeling to determine the set of basis functions that most contribute to the provided dataset. The source separation performance is validated using open-air radar data with injected orthogonal frequency division multiplexed communication signals.