关键词: 协作频谱感知, 凌日搜索优化算法, 卷积神经网络, 协方差矩阵特征值

Abstract:

To address the performance limitations in spectrum sensing with low signal to noise ratio, a cooperative spectrum sensing method based on energy values, the maximum and secondary maximum eigenvalues of the covariance matrix is proposed, which improves the convolutional neural network （CNN） using the transit search optimization algorithm （TSOA） . First, construct the covariance matrix of the received signal and select its maximum and second largest eigenvalues, and a feature vector is constructed by combining the signal energy values, without the prior information of the primary user, to generate the training and test sets of CNN model. The parameters of the CNN are automatically optimized by TSOA, obtaining the optimal CNN model. Then, the test set is used to validate the performance of the TSOA-CNN model. Simulation results show that applying the TSOA-CNN algorithm to multi-dimensional features significantly improves the classification accuracy and spectrum sensing performance of the model.

Key words: cooperative spectrum sensing, transit search optimization algorithm（TSOA）, convolutional neural networks（CNN）, covariance matrix eigenvalue