Research On Compressed Wideband Spectrum Sensing

Spectrum sensing, whose objectives are detecting the presence of primary users(a.k.a licensed users) and finding the idle spectrum bands within license bands, is the precondition for enabling dynamic spectrum access(DSA) considered as an effective way to solve the conflict of the frequency resource scarcity and the increasing demand on available spectrum. With the rapid development of radio technologies and related services, wideband spectrum sensing has recently attracted attentions in the field since it offers more spectrum opportunities in the wideband regime, becoming a crucial trend in spectrum sensing filed. However, the classical Nyquist sampling theorem requires a rate of twice of the highest frequency while sampling the wideband signal, thus such high sampling rate potentially imposes great pressure on the design of analog to digital converter(ADC) and the storage of sampling data. By exploiting the sparse structure of the received wideband signal from the highly low spectrum utilization, compressed wideband spectrum sensing, in fact a combination of compressed sensing theory and wideband spectrum sensing, can effectively reduce the sampling rate and solve a series of problems incurred by the high sampling rate. Despite a quantity of studies in the field of compressed wideband spectrum sensing, many problems are still to be solved. This dissertation focuses on discussing and solving several problems in compressed wideband spectrum sensing and then proposes several effective algorithms for different sensing circumstances. This dissertation contributes in the following aspects:(1) In the research on local compressed wideband spectrum sensing, we analyze the drawbacks of traditional algorithm whose performance is susceptible to many factors such as fading channel and ambient noise. In addition to sparse structure, there actually exist several types of accessible prior information in practical application, such as partially known occupancy status, occupancy probability and fixed spectrum allocation. By exploiting these types of prior information, we propose three compressed wideband spectrum sensing algorithms respectively and simulate their performance in noise-free and noisy case. Simulation results demonstrate that those proposed algorithms can effectively utilize corresponding prior information and significantly improve the performance in spectrum reconstruction and spectrum sensing. Moreover, simulations also discuss the influence of non-ideal prior information and algorithm parameters on proposed algorithms.(2) With respect to compressed wideband spectrum sensing in cooperative networks with multiple sensing nodes, we firstly analyze the drawback of traditional algorithm whose cooperation overhead is extremely high and then propose a cooperative support fusion-based distributed compressed wideband spectrum sensing algorithm by exploiting the joint sparsity structure of signals sensing by different nodes and the large dynamic range property of each signal sensing by single node. The proposed algorithm alternates between local sparse spectrum reconstruction with partially known support and adaptive learning fused support information through local communications among one-hop neighboring nodes. The efficiency of the proposed algorithm is testified by comparing with two existing distributed compressed wideband spectrum sensing algorithms, in terms of the probability of successful reconstruction, detection probability, communication workload and computation burden. Simulation results show that the proposed algorithm can reduce the cooperation overhead while improving the performance. Moreover, the robust performance to the choice of algorithm parameters is also investigated through simulations.(3) In the research on compressed wideband spectrum sensing in cooperative networks with multiple sensing nodes, the existing algorithms can effectively reduce the high signal acquisition costs in wideband spectrum sensing. However, the computation overhead incurred by compressed reconstruction is non-trivial; additional information, such as energy of ambient noise or sparsity order of sparse signal, is required and it is generally unavailable in practice. To address these issues, a novel algorithm, called Karcher mean-based distributed compressed wideband spectrum sensing, is proposed to wideband spectrum sensing in cooperative networks. Our key observation is that the sparse signals are unnecessary to be reconstructed since the task of wideband spectrum sensing interests in the spectrum occupancy status rather than other information of the reconstructed signals. The major novelty of the proposed algorithm relates to the use of Karcher mean as a statistic indicating the spectrum occupancy status, consequently eliminating the compressed reconstruction stage and significantly reducing the computational complexity. Considering limited bandwidth and energy resources per node, a decentralized implementation based on alternating direction method of multipliers is designed to calculate the Karcher mean via one-hop communications only. By comparing with two existing distributed compressed wideband spectrum sensing algorithms, the efficiency of the proposed algorithm is testified in terms of the detection probability of the node network, communication overhead and computational complexity per node. Simulation results show that the proposed algorithm in no need of those additional information can promote the detection performance at low communication overhead and significantly low computation burden. Moreover, impacts of algorithm parameters and system parameters are also investigated through simulations.(4) In complicated electromagnetic circumstance, the number of wideband non-stationary signals becomes large and existing algorithms cannot be applied due to the non-sparsity of the monitored wideband spectrum. By extending the sensing domain from frequency domain to time-frequency domain and exploiting the sparsity in time-frequency domain, a compressed time-frequency domain information sensing algorithm based on short time Fourier transform(STFT)-based is proposed to reconstruct STFT time-frequency domain information from compressed measurements obtained at sub-Nyquist sampling rate. In simulations, the reconstructed performance of typical wideband non-stationary signals is tested, and simulation results show that the proposed algorithm is capable of obtaining STFT time-frequency domain information at relatively low sampling cost.