Multi-channel Cooperative Spectrum Sensing Based on Belief Propagation Algorithm

Multi-channel spectrum sensing is prevailing but also very challenging in wideband cognitive radio systems. Conventional multi-channel spectrum detection such as channel- by-channel scan costs much time and energy. This paper aims to show a novel multi-user cooperative spectrum sensing method which can reduce the sensing ability requirement for secondary users while still guaranteeing the sensing accuracy and effective- ness in a multi-channel cognitive radio context. In our proposed method, each cognitive user chooses an Ideal-Soliton-Distributed number of channels to sense, and the partial detection results are then passed to a confusion center which uses a specially designed Belief Propagation (BP) algorithm to infer the spectrum activities of all the channels. A heuristic method to release the detected channels from the whole spectrum bands is also proposed to reduce the sensing complexity further. Simulation results show that the proposed sensing methods can obtain excellent performance. Additionally in the previous works(4)-(6), cooperation is used to improve the detection performance under multipath, shadowing or noisy radio environment. Here we use it to analyze and infer the spectrum activities of multi-channels based on the cooperation of different secondary users, and to reduce the single-user sensing delay and power consumption. Motivated by the above idea, this paper aims to propose a novel multi-channel cooperative spectrum sensing method which can reduce the sensing ability requirement for secondary users while still guaranteeing the sensing accuracy and effec- tiveness in multi-channel cognitive radio systems. Although it is obvious that the optimal inferring algorithm is MAP algorithm, its complexity is intolerable when the number of channels is relatively large. Fortunately, we find that the resultant inferring model of our method is a typical Bayesian Network. Thus Belief Propagation Algorithm (BPA) will be a more efficient approach for the particular inferring problem. Moreover, long-term spectrum observation shows that most channels are idle during most of the time(1), which makes it very easy and fast to make the decision based on the belief functions of the partial sensing results of the secondary users. However, due to the particular message generating and passing rules of the Bayesian Network, the BPA should be elaborately designed. And because of the similarity between our sensing methods and LT code (7), we introduce the famous Ideal Soliton Distribution to choose the number of sensing channels. In this paper, two kinds of novel wideband spectrum sens- ing methods based on the belief propagation algorithm are proposed, both of which can increase the detection efficiency and reduce the detection delay. For the first sensing method, all the secondary users randomly choose part of channels among the total channels according to the well-known Ideal Soliton Distribution and then pass the sensing results to the fusion center, for example the base station. Afterwards, the fusion center adopts the belief propagation algorithm to infer the activities of all the channels. In the second sensing method, apart from the above process, a strategy is adopted to release the detected channels from the whole spectrum bands to further reduce the sensing complexity.