Research On Spectrum Sensing And Interference Alignment Based On Sensing Methods In Cognitive OFDM System

The spectrum resource, which is non-renewable, has been getting more and more attention for the problem of spectrum scarcity. How to tackle the conflict between spectrum scarcity and spectrum utilization has become the focus of academic research, especially in situation of continuous development of communication services and the wide application of 4G technology. Although using the traditional methods can improve the spectrum utilization to a certain extent, it still can not meet the high demand for the spectrum resources. The emergence of cognitive radio(CR) technology provides a new way to solve this problem.In this paper, I will focus on the spectrum sensing interference alignment based on sensing methods in the orthogonal frequency division multiplexing(OFDM) cognitive radio network. The aim of the paper is to quickly and accurately find free licensed spectrum and reduce the interference when the secondary users share the spectrum. The specific research content mainly includes four parts:(1) In order to solve the problem that the normal spectrum sensing methods could not get the satisfied results and the optimum OFDM spectrum sensing algorithm has no closed form solution and hardly been used in real life, a suboptimal algorithm is proposed based on differential characteristics(DC OFDM). Through the differential operation, the algorithm applies the local optimal solution instead of the global optimal solution to sense the spectrum. Although DC OFDM method gets some loss in performance, it has the closed forms of false alarm probability and detection probability, and reduces the compute complexity, which is more suitable for the practical application. Moreover, by applying the differential operation, the traditional methods which is based on cyclic prefix(CP) and pilot tones(PT) are improved. And the differential characteristics-based cyclic prefix(DCCP) and differential characteristics-based pilot tones(DC-PT) algorithm are proposed.Compared with the original algorithms, the DC-CP and DC-PT algorithms improve the accuracy of the sensing without increasing the computational complexity.(2) In order to solve the effect of timing delay for the traditional spectrum sensing methods, the window function-based cyclic prefix(WF-CP) algorithm is proposed. The window function is used to sum up all the autocorrelation values, which increases the function of CP correlation in the process of spectrum sensing. Moreover, the advantage of WF-CP that it is independent of the timing delay has been proofed both by theory and simulation. On the other hand, to solve the influence of carrier frequency offset(CFO),a new method based on cross correlation matrix of the pilot(CCM-PT) detection method is put forward. The algorithm get the cross correlation value of all the data with the same distribution in the frequency domain, and sum the upper triangular elements of the cross correlation matrix a test statistic. As the result, the robustness of CFO is enhanced,and the sensing performance is improved at the cost of a little increase in computational complexity.(3) In order to solve the problem that the traditional OFDM spectrum sensing algorithms need large amount of prior knowledge and the blind OFDM spectrum sensing methods cost great compute complexity can not get well sensing performance in low signal to noise(SNR), the deep learning method is applied to spectrum sensing for the first time, and the stacked of autoencoders(SAE) blind OFDM spectrum sensing method is proposed in this paper. With the structure feature of deep learning, the useful features of received signal have been extracted automatically and hierarchically, greatly reducing the requirements of priori knowledge, achieving blind spectrum sensing. This method can achieve high sensing performance compared with traditional non-blind spectrum sensing methods in low SNR conditions. Meanwhile, for the sake of solving the draw back of SAE OFDM algorithm that it can not get better sensing performance with high SNR,the joint time domain and frequency domain features SAE(TF-SAE) OFDM and date fusion-based TF-SAE(DF-TF-SAE) are studied. By using the characteristics of the received signal in frequency domain and time domain, these methods make a better sensing performance at the cost of a certain sensing time and hardware consumption.(4) In order to tackle with the problem that the single spectrum sensing algorithm can not adapt to the complex application environment and secondary users share the spectrum will produce interference, the interference alignment method base on spectrum sensing is put forward, according to the proposed spectrum sensing algorithms in this paper and interference alignment technique. According to the results of SNR estimation and the user needs the appropriate spectrum sensing algorithm is selected to achieve a better perception of performance.In order to guarantee the high performance of spectrum sensing,a SNR estimation method based on deep learning(Deep Learning-Based SNR Estimation, DL-SE) is proposed, achieving high estimation accuracy under low SNR situation.On the other hand, IA is applied to share the unused spectrum. The linear finite state Markov chain(LFSMC) and simplified LFSMC(S-LFSMC) IA methods are proposed,respectively. These methods can reduce the dependence of channel state information and increase the user’s signal to interference plus noise ratio(SINR). So, the interference between users is reduced greatly, and the spectrum utilization is improved.