RIP Analysis of Modulated Sampling Schemes for Recovering Spectrally Sparse Signals

Abstract: In this work, we analyze modulated sampling schemes, such as the Nyquist Folding Receiver, which are highly efficient, readily implementable, non-uniform sampling schemes that allows for the blind estimation of a narrow-band signal's spectral content and location in a wide-band environment. This non-uniform sampling, achieved by narrow-band modulation of the RF instantaneous sample rate, results in a frequency domain point spread function that is between the extremes obtained by uniform sampling and totally random sampling. As a result, while still preserving structured aliasing, the modulated sampling scheme is also useful in a compressive sensing (CS) setting. We estimate restricted isometry property (RIP) constants for CS matrices induced by such modulated sampling schemes and use those estimates to determine the amount of sparsity needed for signal recovery. This is followed by a demonstration and analysis of Orthogonal Matching Pursuit's ability to reconstruct signals from noisy non-uniform samples.