随着无线通信技术的不断发展，频谱资源日益紧缺，频谱资源的管理及应用显得十分重要。认知无线电技术通过改善传统的固定频谱分配模式，能有效地提升频谱资源的利用效率。频谱感知是认知无线电的关键技术，随着无线业务逐渐趋于宽带化，传统奈奎斯特速率信号处理方式导致宽带频谱感知过程所需硬件成本较高及信号处理高复杂度等问题。研究宽带压缩频谱感知技术可以降低实际应用对硬件的要求，缓解数据压缩、传输、处理层面的压力。

针对压缩感知中存在重构信号计算复杂度高，及准确性会影响频谱检测结果的问题。提出一种基于协方差矩阵的非重构宽带压缩频谱感知算法。算法将信号经压缩感知得到测量值，进行采样协方差预处理后，将采样协方差矩阵对角线上的值通过预设门限进行频谱判决。仿真结果表明，此算法应用于频谱感知当中具有可行性，能够有效地检测出频谱空洞。

针对传统频谱感知算法的门限限制和低信噪比下检测概率低等问题，提出一种基于粒子群优化最小二乘法支持向量机(PSO-LSSVM)的非重构宽带压缩频谱感知算法。算法将机器学习技术应用到非重构宽带压缩频谱感知当中，通过PSO-LSSVM对压缩感知预处理的测量值数据进行训练分类，得到频谱检测结果。仿真结果表明，所提算法在低信噪比环境下，与传统频谱检测相比频谱检测概率有所提高。

针对单用户频谱感知易受到多径衰落、阴影效应和终端隐匿等因素影响的问题，提出一种基于最大-次最大-最小特征值法(MSMM)的协作频谱感知算法。算法对多用户的测量值数据进行主成分分析以及基于MSMM的特征提取，使用机器学习算法来训练和分类特征值，得到频谱检测结果。仿真结果表明，所提算法在低信噪比环境下，随着协作用户数目的增加，与单用户频谱检测相比，所提的协作频谱感知算法检测概率有着明显改善。

With the continuous development of wireless communication technology, spectrum resources are becoming increasingly scarce, and the management and application of spectrum resources are very important. Cognitive radio technology can effectively improve the utilization efficiency of spectrum resources by improving the traditional fixed spectrum allocation mode. Spectrum sensing is the key technology of cognitive radio. As wireless services gradually become broadband, the traditional Nyquist rate signal processing method leads to the problems of high hardware cost and high complexity of signal processing in the broadband spectrum sensing process. The study of broadband compressed spectrum sensing technology can reduce the hardware requirements for practical applications and relieve the pressure at the data compression, transmission, and processing levels.

To address the problems of high computational complexity of reconstructed signals in compressed sensing, and the accuracy will affect the spectrum detection results. A non-reconstructed broadband compressed spectrum sensing algorithm based on covariance matrix is proposed. The algorithm takes the measured values obtained by compressed sensing, performs sampling covariance preprocessing, and then passes the values on the diagonal of the sampling covariance matrix through the preset threshold for spectrum judgment. The simulation results show that this algorithm is feasible to be applied in spectrum sensing and can effectively detect spectrum voids.

A non-reconstructed broadband compressed spectrum sensing algorithm based on least squares support vector machine algorithm for particle swarm optimization (PSO-LSSVM) is proposed to address the problems of threshold limitation and low detection probability under low signal-to-noise ratio of traditional spectrum sensing algorithms. The algorithm applies machine learning techniques to non-reconstructed broadband compressed spectrum sensing, and the PSO-LSSVM is used to train and classify the measurement data pre-processed by compressed sensing to obtain the spectrum detection results. The simulation results show that the proposed algorithm improves the probability of spectrum detection compared with traditional spectrum detection in low SNR environment.

A cooperative spectrum sensing algorithm based on the maximum-second maximum-minimum eigenvalue method (MSMM) is proposed for single-user spectrum sensing that is susceptible to multipath fading, shadowing effects, and terminal concealment. The algorithm performs principal component analysis and MSMM-based feature extraction on the measured value data of multiple users, and uses a machine learning algorithm to train and classify the feature values to obtain spectrum detection results. The simulation results show that the detection probability of the proposed cooperative spectrum sensing algorithm is obviously improved compared with single-user spectrum detection with the increase of the number of cooperative users in the low SNR environment.