Precolouring in compressive spectrum estimation for cognitive radio

One of the major challenges in cognitive radio (CR) networks is the need to sample signals as efficiently as possible without incurring the loss of vital information. Compressive Sensing (CS) is a new sampling paradigm which provides a theoretical framework for sub-sampling signals which are characterized as being sparse in the frequency domain. The random demodulator (RD) is a CS-based architecture which has been employed to acquire frequency sparse, bandlimited signals which typify the signals which often occur in many CR-related applications. This paper investigates the impact of precolouring upon CS performance by combining the RD with an autoregressive (AR) filter model to enhance compressive spectral estimation. Quantitative results with quadrature phased shift keying (QPSK) modulated multiband signals, corroborate that adopting a precolouring strategy both reduces the spectral leakage in the power spectrum, and concomitantly improves the overall signal-to-noise ratio (SNR) performance of the compressive spectrum estimator.