利用卷积神经网络和协方差的协作频谱感知算法

针对基于信号协方差矩阵的频谱感知算法门限难于准确得到及没有充分利用原始信号信息等问题,提出了基于卷积神经网络和协方差矩阵的协作频谱感知算法。首先将接收的I、Q两路正交信号的归一化协方差矩阵组成双通道输入矩阵,然后使用卷积神经网络直接提取协方差矩阵的特征信息,并进行训练得到分类器,最后使用训练好的模型进行频谱感知。仿真结果表明,本文所提出的频谱感知算法优于对比算法,在信噪比为-13 dB、40个次用户协作感知时,本文算法虚警概率低于0.1,检测概率达到0.9以上。 

Cooperative Spectrum Sensing Using Convolutional Neural Networks and Covariance

A cooperative spectrum sensing algorithm based on convolutional neural network and covariance matrix is proposed to solve the problems that the threshold value of spectrum sensing algorithm based on signal covariance matrix is difficult to get accurately and the raw signal information is not fully utilized. Firstly, the normalized covariance matrix of the received I and Q orthogonal signals is formed into a two-channel input matrix. Then, the characteristic information of the covariance matrix is extracted by using convolutional neural network to train the classifier. Finally, the trained model is used for spectrum sensing. The simulation results show that the proposed spectrum sensing algorithm is superior to the contrast algorithms. The false alarm probability of this algorithm is less than 0.1 and the detection probability is above 0.9,when the SNR is -13 dB and 40 secondary users cooperate in sensing.