基于累积和的改进超宽带循环平稳检测算法

针对超宽带循环平稳检测存在的门限难以设定、低信噪比下检测延迟较大的问题，提出了基于累积和的改进超宽带循环平稳检测算法。首先将信号整个三维循环谱归一化为二维灰度图，与噪声对应的灰度图比较差异，再将两类图像放入卷积神经网络（Convolutional Neural Network,CNN）自行训练提取特征，解决门限难以确定的问题。若分析三维循环谱的时间块长过短，将导致信号灰度图特征在有无噪声情况下区别不大；若块长过长会导致检测延迟较大。为此，采用累积和算法提取网络全连接层输出的信号概率作为累积和的观测统计量，自适应检测所需采样时间长度。将所提算法与传统循环平稳检测以及结合了CNN的循环平稳检测进行对比，仿真表明所提算法在低信噪比下性能最优。

An Improved UWB Cycliostationary Detection Algorithm Based on Cumulative Sum

In order to solve the problem that the threshold of ultra-wideband(UWB) cyclostationarity detection is difficult to be set and the detection delay is large at low signal-to-noise ratio(SNR),an improved UWB cyclostationarity detection algorithm based on cumulative sum(CUSUM) is proposed.Firstly,the whole three-dimensional(3D) cyclic spectrum of the signal is normalized into a two-dimensional(2D) grayscale map,which is different from the grayscale image of the 3D cyclic spectrum of the noise.Then,the two images are put into a convolutional neural network(CNN) to train and extract features,so as to solve the problem that the threshold is difficult to be set.If the time block of the 3D cyclic spectrum is too short,it will lead to little difference in the characteristics of the gray image of the signal with or without noise.If the block is too long,it will lead to a large detection delay.Therefore,the signal probability output from the network''s full connection layer is extracted as the observation statistics of CUSUM,and the sampling time length required for adaptive detection is calculated.The proposed algorithm is compared with the traditional cyclostationary detection and the cyclostationary detection combined with CNN.The simulation results show that the proposed algorithm has the best performance at low snr.