基于深度卷积神经网络的协作频谱感知方法

针对传统卷积神经网络(CNN)频谱感知方法提取特征能力受限于网络结构简单，增加网络结构又容易出现梯度消失等问题，该文通过在传统卷积神经网络中添加捷径连接，实现输入层恒等映射更深的网络，提出一种基于深度卷积神经网络(DCNN)的协作频谱感知方法。该方法将频谱感知问题转化为图像二分类问题，对正交相移键控(QPSK)信号的协方差矩阵进行归一化灰度处理，并作为深度卷积神经网络的输入，通过残差学习训练深度卷积神经网络模型，提取2维灰度图像的深层特征，将测试数据输入到训练好的模型中，完成基于图像分类的频谱感知。实验结果表明:与传统的频谱感知方法相比，在低信噪比(SNR)下、多用户协作感知时，所提方法具有更高的检测概率和更低的虚警概率。

Cooperative Spectrum Sensing Method Based on Deep Convolutional Neural Network

The traditional spectrum sensing method of Convolutional Neural Network (CNN) has a simple network structure which limits the ability of feature extraction. To solve the problem of gradient disappearance, a cooperative spectrum sensing method based on Deep Convolutional Neural Network (DCNN) is proposed in this paper, in which shortcut connections are added to the CNN to realize the deeper network of input level identity radiation. This method transforms the spectrum sensing problem into the image binary classification problem, and performs normalized gray level processing on the covariance matrix of Quadrature Phase Shift Keying (QPSK) signal as the input of DCNN, which trains DCNN model through residual learning and extracts the deep image features of the two-dimensional grayscale image. The testing data is input into the trained model and spectrum sensing based on image classification is completed. The experimental results show that the proposed method has higher detection probability and lower false alarm probability than the traditional spectrum sensing method when the Signal to Noise Ratio (SNR) is low and multiple users collaborate in sensing.