Research On Spectrum Sensing Method Based On Optimized Clustering Under Alpha Stable Distribution Noise

The fixed spectrum resource allocation mode makes most of the authorized frequency bands not fully utilized and the spectrum utilization rate is low.Cognitive radio technology can dynamically realize the environment awareness in real time,which is an effective way to solve the low utilization of spectrum resources.Existing spectrum sensing techniques often use Gaussian noise modeling,but the actual sensing environment is often disturbed by nonGaussian noise with spike pulse characteristics,resulting in the performance of detection methods based on Gaussian noise modeling is affected under non-Gaussian noise.Alpha stable distribution noise in non-Gaussian noise can be a good fit for the noise caused by human factors and natural factors,and spectrum sensing can be regarded as a dichotomous classification problem,the clustering algorithm in machine learning can achieve good classification of data.Therefore,in this paper,we combine the clustering algorithm in machine learning to conduct an in-depth study on the spectrum sensing technique under Alpha stable distribution noise.Firstly,for the problem that the judgment threshold value is difficult to determine and complicated to estimate in the single-antenna low SNR environment,an improved viscous bacteria-based optimal clustering spectrum sensing method is proposed.The method uses normalized fraction low order and IQ signal decomposition to suppress the excessive noisy sample values caused by Alpha stable distribution noise trailing phenomenon.On this basis,a clustering algorithm is used to classify the spectrum states,and a Fuzzy c-means(FCM)clustering algorithm based on viscous optimization is proposed to overcome the randomness of the initial center of the FCM algorithm and obtain a stable initial clustering center to further improve the sensing performance.The scheme reduces the reliance on a priori information,avoids complex threshold estimation,and effectively improves the single-antenna spectrum sensing performance under low signal-to-noise ratio conditions.Second,to address the problem of poor sensing performance due to small differences in covariance matrix elements between primary users of different states of the multi-antenna system,an improved gray wolf-based optimal clustering spectrum sensing method is proposed.The method uses quadratic fractional low-order covariance matrices to expand the differences between matrices;the fusion grouping method is used to construct two-dimensional feature vectors to reduce the computational complexity caused by signal decomposition.An improved gray wolf optimization algorithm is proposed to overcome the randomness of the initial prime of the k-means algorithm and obtain a stable prime to further improve the clustering accuracy,thus improving the perception performance.The simulation shows that the method can further improve the multi-antenna spectrum sensing performance under Alpha stable distribution noise.Finally,to address the problem of poor sensing performance due to the interference of SU users by abnormal users in the actual sensing environment.An anomalous environment spectrum sensing method based on improved dung beetle optimized clustering is proposed.The method uses the LU matrix to enhance the features of sample data,the median data fusion method is used to reduce the interference of anomalous users to sensing users.The improved dung beetle optimization algorithm is proposed to overcome the strong dependence of kmedoids on initial points and obtain stable initial points to further improve the clustering accuracy when anomalous users interfere,thus improving the sensing performance.In the simulation part,it is verified that the proposed method can reduce the interference of anomalous users to improve the perceptual performance.