Spectrum Sensing Algorithms For Cognitive Radio Networks

With the proliferation of wireless services, the advent of new high data rate wireless applications, the rapid deployment of new wireless devices, and the compelling need for mobile Internet access, demand for the additional spectrum/bandwidth is dramatically increasing in the last decade and is expected to grow even more rapidly in the coming years. As an inherently limited natural resource, the access to spectrum is regulated by the government agencies. Within the conventional static spectrum allocation management, the majority of the frequency bands have been exclusively licensed to the primary users where unlicensed access to the spectrum is prohibited. It has become exceedingly hard to allocate vacant bands to deploy new services. This spectrum limitation has had profound impacts on the research directions of the wireless communications community. However, it has been shown that the current static spectrum allocation and license management has resulted in inefficient utilization of this limited resource across time and frequency. That is, the current spectrum scarcity is mainly due to the inefficient static spectrum regulations rather than the physical shortage of the spectrum. Therefore, regulatory bodies and researches across the world are contemplating more dynamic spectrum access policies and innovative communication technology to exploit the wireless spectrum in a more intelligent and flexible way to resolve the spectrum scarcity.Cognitive radio, first proposed by Mitola, is a new concept in wireless communications and is a promising technology to solve the spectral crowding problem by introducing the opportunistic usage of the primary user's licensed frequency bands that are not occupied by the primary users. It is the key enabling technology that enables next generation communication networks, also known as dynamic spectrum access (DSA) networks. As "the next big bang" in communications, cognitive radio has been receiving an increasing attention and attracting significant interest within the wireless communication research community. The basic idea of cognitive radio is spectrum sharing. It allows the secondary users exploit the primary users'licensed spectrum resources that are not currently occupied by the primary users. Therefore secondary users need to perform spectrum sensing to find the available spectrum resources. Once the primary user reclaims the spectrum usage right, the secondary user should vacate the channel so that no harmful interference is generated due to secondary users' transmission. Therefore, spectrum sensing and analysis is the first critical step towards dynamic spectrum management and the core technology for the application of cognitive radio.Recent years there have been numerous conferences, workshops and sessions specifically on cognitive radio, as well as many special issues of the prestigious journals. There been many significant developments on the spectrum sensing research for the cognitive radio. And quite a few algorithms have been proposed in different scenarios during the past few years, such as the matched filter detector, the cyclostationarity models, the energy detector, as well as the cooperative sensing. However, most spectrum sensing focus on the situation of Gaussian noise, while in practice the noise is known to be non-Gaussian due to the impulsive phenomena in some situation. These detectors'performance will be degraded substantially in the presence of non-Gaussian noise. Also due to the effect of shadowing, fading and noise uncertainty, exact prior knowledge of the primary user's signal is hard to acquire for the secondary user, besides that primary user'SNR observed at the receiver of secondary user is quite low and fluctuating over a large region. In order to improve the capacity of cognitive radio networks, secondary user need to achieve a satisfying detection performance within a short sensing time. Furthermore, cooperative sensing results in extra traffic overhead and serious energy consumption. So spectrum sensing in cognitive radio still face great challenges, and great efforts and in-depth studies are required to address these issues to promote the development of cognitive radio and its application.This thesis surveys recent advances in research related to cognitive radios and mainly focus on the spectrum sensing issues. First, we introduce the background, fundamentals of cognitive radio technology and architecture of a cognitive radio network in Chapter?. Existing works in spectrum sensing, different types of detection techniques, and cooperative spectrum sensing are reviewed in Chapter?, as well as the spectrum sensing challenges. Then we propose several effective spectrum sensing detectors with different prior knowledge in different situations to solve the afore-mentioned problems. The main contributions of our work can be summarized as follows:1) Accurate and Efficient Energy Detector in Complex Gaussian Noise with Perfect Noise Variance Information:Energy detector is a very useful detector for spectrum sensing in cognitive radio systems and has been widely employed for sensing optimization. However, there is certain mismatch between the theoretical and actual detection performance of the conventional energy detector in complex Gaussian noise, which will undermine the optimization accuracy. To overcome this problem, we propose an accurate and efficient energy detector for the spectrum sensing in complex Gaussian noise in Chapter?and derive its theoretical detection performance expressions based on the approximated distributions of its test statistic under different hypothesis. Numerical results demonstrate the efficiency of the proposed energy detector and its superiority to the conventional one. That is, its theoretical detection performance can match well with the actual detection performance, which is of great importance for the design of optimal spectrum sensing in cognitive radio systems. Also the proposed energy detector can achieve a higher detection probability than the conventional one, which means better spectrum utilization efficiency and lower interference to the primary user.2) Robust and Blind Sample Covariance Matrix Eigenvalue Based Spectrum Sensing Detecor:When accurate noise variance is known, energy detector is optimal, however, its performance is sensitive to the noise uncertainty. To achieve a satisfying detection performance when no prior noise information is available, we propose a robust and blind eigenvalue-based spectrum sensing detector in Chapter IV. The eigenvalue-based detector does not require any prior knowledge of the primary user's signal or the noise. Its test statistic is the ratio of the largest to the smallest eigenvalues of the sample covariance matrix. The robustness of the detector can be achieved by normalizing the observed signal vectors. After the normalization, the distributions of its test statistic under different hypotheses become more distinct, which will enhance the detection performance significantly. The threshold is set according to latest findings about the statistics of the eigenvalues in the random matrix theory. Simulations in different scenarios confirm the superiority of the proposed detector. Specifically, the proposed detector offers a much better detection performance over other detectors in the situation of non-Gaussian noise. And compared with the existing detector, the performance loss in Gaussian noise is just about 0.5dB which is almost negligible.3) Fast and Robust Spectrum Sensing Exploiting a Small Number of Noise-only Samples:To achieve a satisfying detection performance in a short sensing time, Chapter V proposes a fast and robust spectrum sensing detector based on the Kolmogorov-Smirnov (K-S) test. The K-S detector needs only a short sequence of noise samples, which is essentially the same requirement as the energy detector. The basic procedure involves computing the empirical cumulative distribution function of some decision statistic obtained from the received signal, and comparing it with the ECDF of the channel noise samples. The largest absolute value of the difference between these ECDFs is used as the test statistic to make the sensing decision by comparing it with a threshold decided by the target false alarm probability. Extensive simulations demonstrate that the proposed K-S detectors offer superior detection performance and faster detection. It significantly outperforms the existing detectors in the presence of non-Gaussian noise. In Gaussian noise, it is less sensitive to the noise uncertainty than the energy detector and requires much less sensing time than the eigenvalue-based detector.4) Robust Sequential Spectrum Sensing with a Prior Noise Knowledge:When the region of the primary user's SNR is more dispersed, sequential detectors can take into the account of the actual value of the observed signal to make a quicker sensing decision. That is when primary user's instantaneous SNR is high, secondary user can make a right decision only with a small number of observations. Consequently, the sensing consumption and the interference to the primary user can be decreased. Given the advantage of sequential sensing, we propose two robust sequential K-S test based spectrum sensing detectors with different prior noise knowledge in Chapter VI. The proposed sequential detectors can enhance the sensing agility without the prior knowledge of the statistics of the primary user's signal. Simulations confirm the efficiency and advantage of our proposed sequential detectors over other existing sequential detectors. Specially, the proposed detectors offer great performance improvement in the situation of non-Gaussian noise, where the unavailable distribution function of the mixture of noise and signal leads to the failure of existing sequential detectors. When primary user'SNR region is large, the proposed sequential detectors can decrease the sensing cost and achieve a fast sensing decision.As the last part of the dissertation, Chapter?summarries our work and analyzes the prospect of the developing tendency aiming to further improve the system performance of cognitive radio, such as sensing design in correlated noise, cooperative sensing optimization, as well as the cross-layer joint sensing optimization.