Research On HMM-based Spectrum Sensing In Cognitive Radio Networks

With explosive growth of wireless services and the number of mobile devices in recent years,the limited radio spectrum has become more and more scarce.Most allocated licensed bands are severely underutilized and cannot been reused by other users in traditional spectrum management policy,which makes the radio spectrum difficult to meet the increasing demand of the wireless services.Cognitive radio(CR)technology allows the secondary user to access the band of a primary user as long as the primary user is in the absence state or the secondary user causes tolerable interference to the primary user.Obviously,it can improve the spectrum utilization and alleviate the spectrum scarcity problem.Thus,once the concept of CR was proposed,it immediately attracted much attention in communication field and industry,and there have been lot of research works on it.In CR networks,the secondary user must sense the status of the primary channel before transmission.Therefore,spectrum sensing is one of the key technologies and the cornerstone of cognitive radio.Traditional spectrum sensing algorithms based on three assumptions,which are given as follows.First,the primary user only has two states,namely,the absence or the presence with a constant transmit power.Second,the secondary user must sense the primary channel at every slot and decide the transmit strategy according to the spectrum sensing result.Third,the secondary user has unlimited energy supply.In this paper,we focus on the extension of the three assumptions mentioned above.The main contents and contributions are summarized as follows.Firstly,the spectrum sensing algorithm in the scenario where the primary user has more than one transmit power levels is studied.A continuous hidden Markov model(CHMM)based blind spectrum sensing algorithm for not only detecting the presence of primary user but also recognizing the transmit power level of the primary user is proposed.By combing the wavelet singularity detection with k-means clustering algorithm,an effective method for estimating of the number of primary transmit power levels is proposed and the training problem of CHMM is solved.Two different spectrum sensing strategies are designed according to the actual demand of users.Simulation results show the proposed algorithm can effectively sense the true state of the primary user,and give the applicable scenario of the two sensing strategies,respectively.Secondly,the optimization problem for finding the optimal spectrum sensing interval when the primary channel occupancy states form a Markov process and the secondary user has unlimited energy supply is studied.A discrete Hidden Markov model(DHMM)is used to describe the imperfect spectrum sensing process over Rayleigh fading channels.Based on the sensing results,a hybrid interweave/underlay mode is exploited by the secondary user.The optimal spectrum sensing interval of the secondary user is obtained by solving the tradeoff problem among the average energy consumption for spectrum sensing,the average throughput of the secondary user,the average decreased throughput of the secondary user and the average interference to the primary user.Simulation results verify our analysis and show that our proposed algorithm is advantageous in the average energy consumption for spectrum sensing and the average throughput of the secondary user.Thirdly,the optimization problem for finding the optimal spectrum sensing interval when the primary channel occupancy states form a Markov process and the secondary user has limited energy supply is studied.A DHMM is used to describe the imperfect spectrum sensing process.Energy harvesting(EH)technique is used to sustain the operation of the secondary user.A partially observable Markov decision process(POMDP)framework is used to determine the optimal spectrum sensing interval to explore the best trade-off between the average throughput of the secondary user and the average interference caused to the primary user.A myopic spectrum sensing interval optimization policy is proposed to reduce the computation complexity.The precondition of using the Markov decision process(MDP)framework is analyzed and the detailed procedure is given.Simulation results verify the efficiency of the proposed algorithm and show that,compared with the traditional spectrum scheme having fixed spectrum sensing interval,the proposed algorithm substantial increases the average throughput of the secondary user,at the cost of the small interference caused to the primary user.Finally,the joint optimization of spectrum sensing energy,trigger energy and spectrum sensing interval to maximize the whole satisfaction degree(WSD)of CR network,when the primary channel occupancy states form a Markov process and the secondary user has limited energy supply is studied.A DHMM is used to describe the imperfect spectrum sensing process.The WSD of CR network is defined as a function of the average throughput of the secondary user and the interference caused to the primary user.The WSD maximization problem is formulated as a mixed integer non-linear programming problem(MINLP).A differential evolution(DE)algorithm is used to optimize the spectrum sensing energy,trigger energy and spectrum sensing interval.Simulation results validate the correctness and effectiveness of the proposed algorithm.