TFR Reconstruction From Incomplete m-D Signal via Adaptive Hadamard Product Parametrization

In micromotion signature analysis, the radar return signal with missing sampling may cause defocused time-frequency representation (TFR) and thus prevent micromotion characteristics acquisition. To address the issue, we present an adaptive time-frequency distribution reconstruction method based on L-1 regularization. First, the L-1 regularization is expressed as a combination of two L-2 regularizations based on Hadamard product parametrization. Then, the iterated Tikhonov regularization is applied to solve each L-2 regularization alternatively. Moreover, the regularization parameter is updated adaptively based on the matching pursuit principle at each iteration. Finally, the reconstructed TFR is updated based on least-square-error criterion to eliminate the attenuation of signal amplitude. Simulation and measurement data examples have demonstrated the effectiveness of the method.