Cooperative Spectrum Sensing with Covariance Matrix Decomposition and CNN
According to the underutilization of main characteristic in signal sampling covariance matrix and an unstable detection threshold,a cooperative spectrum sensing method is proposed by covariance matrix decomposition of convolutional neural networks(CNN).Firstly,the covariance matrix of the received signal sampling matrix is Cholesky decomposed and the statistics are calculated to fully extract the characteristics of the two signals.Secondly,the statistics obtained by multiple secondary users are accumulated to form a statistical characteristic matrix for improving both collaboration and detection accuracy.Finally,by using the feature extraction ability of CNN to high-dimensional matrix,the CNN classification model is trained by the training dataset,and the spectrum results are predicted by the test dataset.Experimental results show that the proposed algorithm has higher training accuracy and shorter experimental time compared with SVM,traditional CNN and other algorithms.Given false alarm probability of 0.1,the detection probability of the proposed scheme outperforms the traditional CNN and SVM by 60％and 69％at SNR of-15 dB,respectively.
NETL
NSTL
万方数据
