| With the rapid increase of radio users and the increasing scarcity of spectrum resources,to utilize the spectrum resources more efficiently,many researchers focus on the wideband spectrum sensing(WSS)technology.However,the WSS technology based on Nyquist sampling is confronted with the problem of high sampling rate,and the WSS technology based on compressed sensing can obtain the spectrum by Sub-Nyquist sampling,which reduces the sampling rate of the radio frequency receiver front-end.Nevertheless,there are still many problems for the WSS technology based on compressed sensing in practice.Firstly,the measurement matrix must be decided before applying compressed sensing technology,which determines the capacity of the compression performance.Although the random measurement matrix has better compression performance,it faces the problem of too large memory in practice.Secondly,when the sparsity of signal is unknown,the user has to choose the conservative number of compressive measurements,which not only increases the computational complexity,but also undermines the advantage of system in reducing sampling rate.In addition,the complex environment will degrade the detection performance of WSS,which making the single user cannot obtain the accurate detection results.To solve the problems above,this dissertation studies the construction and optimization of the measurement matrix,adaptive compressive spectrum sensing with sparsity unknown,adaptive compressive spectrum sensing in dynamic environments,and the problem of poor detection performance in low signal to noise ratio(SNR).The main work contents and innovative achievements of this dissertation are as follows:1.The optimization and performance validation of chaotic measurement matrix.The maximum compression efficiency is determined by the measurement matrix in compressed sensing.To overcome the difficulty to verify the restricted isometry property,the Lyapunov exponent is adopted instead of the restricted isometry property to verify the performance of measurement matrix,and there is no need to consider the effection of sparsity when calculating the Lyapunov exponent,which can decrease the computational complexity and suit for any size of matrix.In order to improve the performance of measurement matrix,using the segmented method,the movement law of the chaotic system should be more complicated,which can improve the performance of chaos.The randomness and autocorrelation of four-section Tent map are used to verify the effectiveness of segment method.Compared with other measurement matrices,all elements of four-section Tent measurement matrix are generated by only two parameters,and its performance is close to Gaussian measurement matrix,which verifies its advancements.2.Research on adaptive compressive spectrum sensing technology with sparsity unknown.Sparsity unknown may lead to large size of measurement matrix,which will reduce the compression efficiency.To solve this problem,the two step adaptive compressive spectrum sensing method(TS-ACSS)is proposed.In first step,by analyzing the relationship between the inner product of compressive measurements and the sparsity,a sparsity estimation method based on curve fitting for reference variables is proposed to choose the number of compressive measurements preliminarily.Based on this result,an adaptive spectrum sensing method is proposed in second step.According to the iteration interval,the number of measurements is increased gradually to reconstruct the spectrum,until the stop criterion is satisfied.The sparsity correction result is obtained and the final number of compressive measurements is very close to the actual requirement,which largely reduces the number of compressive measurements required by the traditional compressive spectrum sensing method.Finally,the performance of TS-ACSS algorithm is verified by the sparsity estimation,the iteration number,the compression ratio and other criterions.Compared with the traditional spectrum sensing algorithms based on the othogonal matching pursuit method and the successive iteration algorithm,TS-ACSS algorithm can avoid excessive measurements while using the orthogonal matching pursuit method,and its iteration number is far less than the successive iteration algorithm;3.The optimization algorithm of TS-ACSS in dynamic environment.To solve the problem of requiring too many observations and being poor adaptability for dynamic environment in the TS-ACSS algorithm,the TS-ACSS algorithm is improved in three aspects.1)To perform the pre-decision before running the TS-ACSS algorithm,if the channel state is not changed,the algorithm is stopped,which saves the computing resources and reduces the sensing time.2)The first step of the TS-ACSS algorithm is optimized.Using the former sensing result to update the threshold,the accuracy of sparsity estimation can be improved,and only one successful observation is required for sparsity estimation.3)In the second step of the algorithm,the stop criterion based on binary sensing results is used,which make the spectrum recovery error be lower.Through three aspects of improvements,the sensing performance of TS-ACSS algorithm is improved in dynamic environment.Finally,the simulations and experiments show that the reconstruction error of the TS-ACSS optimization algorithm is 20%lower than the TS-ACSS algorithm,and the estimation performance of TS-ACSS optimization algorithm is more stable;4.Research on the cooperative spectrum sensing technology in low SNR.To solve the problem that the sensing performance of single node is poor in low SNR,a two bits mixed combination method based on the recovery order is proposed.First of all,the relationship between the recovery order in orthogonal matching pursuit algorithm and the reliability of the channel sense results is analyzed,then the different weights are assigned according to the channel sense results,which fully utilizes the relative position between the primary user(PU)and the secondary user(SU).In order to save the spectrum resources for transmitting decisions,the channel occupancy state and its corresponding weights are represented by two bits information.According to the recovery order,three different weights are assigned to different subchannel sensing results.In terms of the detection rate and false alarm rate,the cases of insufficient sampling rate and low SNR are discussed in simulations,the results show that the two bits mixed combination method is valid in low SNR and Or Rule is the best decision fusion method while the sampling rate is insufficient.In addition,the two bits mixed combination method is compared with Or Rule,in the condition of achieving the same detection rate,the former method reduces the false alarm rate by 30%,which verifies its advancement. |
