Research On Spectrum Sensing Forcognitive Radio Based On Machine Learning And Compressive Sensing

The dramatic increase in wireless access and low spectrum utilization has became a paradox due to the conventional spctrum allocation policy. Cognitive Radio is the key technology to solve this paradox. Cognitive radio can sense the electromagnetic environment to choose the spectrum hole which is called spectrum sensing. Through spectrum access, CRN can improve the spectrum utilization. However spectrum sensing faces some major challenges such as: SNR wall; high hardware cost for widebind spectrum sensing; node failure for single CR uer. This paper will consider the above three problems in our research.Firstly, a modular spectrum sensing method based on machine learning is proposed. one drawback of the traditional energy detection is that ifthe SNR of received signal is low and the number of the received signal samples is small the corresponding performance may decreaseseriously. In order to overcome this drawback a novel method with the purpose of obtaining a nonlinear threshold for energy detection based on machine learning is proposed in this paper. The proposed method focuses on one singlepoint and one antenna scenario, which can be divided into Offline module and Online module. In the Offline module, the decision function with probability of false alarm will be obtained. In the Online module, the decision functions obtained in the Offline module are used as the nonlinear thresholds for energy detection to verify if the primary user ispresent. The experimental results show that the receiver operating characteristic (ROC) curve of proposed approach is much better than traditional energy detection.Secondly, a wideband spectrum sensing method is proposed. To solve the problem of signal sampling and the reconstruction, an improved Hadamard measurement matrix based on Walsh code and the reconstruction algorithm based on RVM are proposed. The simulation results show that we can use20%Nyquist Sampling rate to obtain a high performance spectrum sensing.Last, we proposed a cooperative wideband spectrum sensing based on multi task compressive sensing system. In wideband cognitive radio (CR) networks, spectrumsensing is an essential task for enabling dynamic spectrum sharing,but entails several major technical challenges: very high samplingrates required for wideband processing, limited power and com-puting resources per CR, frequency-selective wireless fading, andinterference due to signal leakage from other coexisting CRs. A cooperative approach to wideband spectrum sensingis developed to overcome these challenges. To effectively reduce thedata acquisition costs, a compressive sampling mechanism with Walsh measurement matrix is utilized which exploits the signal sparsity induced by network spectrum under-utilization. To collect spatial diversity against wirelessfading, multiple CRs collaborate during the sensing task by multi task compressive sensing,which is derived to attain high sensing performance at a reasonable computational costand power overhead. To identify spurious spectral estimates dueto interfering CRs. Simulations testify the effectiveness ofthe proposed cooperative sensing approach in CR networks.