Sinusoidal frequency modulation sparse recovery for radar micro-Doppler analysis

This paper derives a new approach, named sinusoidal frequency modulation sparse recovery (SFMSR) for micro-Doppler (m-D) analysis, by exploiting the micro motion spectrum sparsity in SFM signal space. SFMSR employ the Fourier modulation dictionary, which is determined only by the “frequency in SFM signal space”. Unlike other SR-based methods whose dictionary is discretization of a 3D space parameterized by the micro motion amplitude, frequency and initial phase, the SFMSR reduce the m-D analysis to 1D parameter optimization, and therefore can enhance the precision, stability and computational efficiency simultaneously. The temporally correlated sparse Bayesian learning (TSBL) in SFM signal space is employed to decompose the signal and produce highly sparse solutions. Simulation results indicate that the proposed method outperforms the existing methods in accuracy and robustness.