Penalized L1 Minimization for Reconstruction of Time-Varying Sparse Signals
Most practical signals of interest have a sparse representation if they are transformed into a suitable basis. The transformed signal can then be compressed, but this wastes resources as most of the sampled information is discarded after compression. Compressive Sensing (CS) leverages the compressibility of the signal by directly acquiring a smaller quantity of random linear measurements that contain a little redundancy in the information level. Thus, CS appears to be an excellent approach for applications in which data acquisition is expensive such as imaging at non-visible wavelengths and sampling made by wireless sensor nodes.