Compressive Sensing of Multiband Spectrum towards Real-World Wideband Applications.

PhD Theses.Spectrum scarcity is a major challenge in wireless communication systems with their rapid evolutions towards more capacity and bandwidth. The fact that the real-world spectrum, as a nite resource, is sparsely utilized in certain bands spurs the proposal of spectrum sharing. In wideband scenarios, accurate real-time spectrum sensing, as an enabler of spectrum sharing, can become ine cient as it naturally requires the sampling rate of the analog-to-digital conversion to exceed the Nyquist rate, which is resourcecostly and energy-consuming. Compressive sensing techniques have been applied in wideband spectrum sensing to achieve sub-Nyquist-rate sampling of frequency sparse signals to alleviate such burdens. A major challenge of compressive spectrum sensing (CSS) is the complexity of the sparse recovery algorithm. Greedy algorithms achieve sparse recovery with low complexity but the required prior knowledge of the signal sparsity. A practical spectrum sparsity estimation scheme is proposed. Furthermore, the dimension of the sparse recovery problem is proposed to be reduced, which further reduces the complexity and achieves signal denoising that promotes recovery delity. The robust detection of incumbent radio is also a fundamental problem of CSS. To address the energy detection problem in CSS, the spectrum statistics of the recovered signals are investigated and a practical threshold adaption scheme for energy detection is proposed. Moreover, it is of particular interest to seek the challenges and opportunities to implement real-world CSS for systems with large bandwidth. Initial research on the practical issues towards the real-world realization of wideband CSS system based on the multicoset sampler architecture is presented. In all, this thesis provides insights into two critical challenges - low-complexity sparse recovery and robust energy detection - in the general CSS context, while also looks into some particular issues towards the real-world CSS implementation based on the i multicoset sampler