Abstract:

For short pulses, it has been verified that the sub-Nyquist sampling system based on Gabor frames with exponential reproducing windows holds nice performance, but when the frames are highly redundant, the traditional methods for subspace detection may fail for have large errors. Firstly, we design the dual Gabor dictionaries, dividing it to blocks. Then the reconstruction model is built under the blocked dictionaries. Consequently, we introduce the signal space projection into the reconstruction model and propose simultaneous compressive sensing matching pursuit (SCoSaMP) based on block Gabor dictionaries signal space projection, with the converge restricted conditions analyzed. Additionally, the error bound based on oracle estimation is deduced when the sampling system is contaminated by noise, and the denoising method is analyzed. Finally, simulation experiments prove that the new method improves the recovery accuracy, decreases the channel number and enforces the robustness of the sampling system compared with the traditional methods.