Block-Structured Compressed Spectrum Sensing With Gaussian Mixture Noise Distribution

Abstract

In real cognitive radio systems, the additive noise does not always follow Gaussian distribution, e.g., man-made noise and impulse noise. This letter addresses the problem of sparse spectrum sensing when the additive noise follows Gaussian mixture distribution. What is more, the key parameter of the distribution is unknown. By introducing a prior variable that can control the block sparse structure of the solution, the prior knowledge of the block structure of the signal of the primary users is effectively explored. All of the unknown variables are calculated in iterative manner with closed form based on the theory of vibrational message passing. Simulation results show that the proposed algorithm is robust to Gaussian mixture noise and can obtain better performance.