基于小波变换和残差神经网络的全盲频谱感知方法

为了解决非平稳信号的时频图特征难以提取的问题，提出了基于小波变换和残差神经网络的全盲频谱感知方法。该方法通过连续小波变换将捕获的主用户信号变换成时频信息矩阵，同时转化为图片作为输入，通过残差网络进行训练和识别。仿真测试了不同小波基对非平稳信号的分解能力和所提算法在各种复杂无线信道环境下对非平稳信号的检测性能和泛化能力，以及对不同主用户信号的适应能力。结果表明，在信噪比为-16 dB时，该方法能在虚警概率为0.1时达到0.92的检测概率，同时amor小波更适合用于非平稳信号的分解且训练识别能力更优。

A full-blind spectrum sensing approach based on wavelet transformation and residual neural network

In order to solve the problem that it is difficult to extract the time-frequency image features of non-stationary signals,a full-blind spectrum sensing method based on wavelet transform and residual neural network is proposed.In this method,the captured main user signal is transformed into a time-frequency information matrix by continuous wavelet transform.Then the matrix is treated as input picture,which is trained and identified through the residual network.The decomposition ability of different wavelet bases on non-stationary signals and the detection performance and generalization ability of the proposed algorithm on non-stationary signals in various complex wireless channel environments are simulated,as well as the adaptability of this method to different primary user signals.The results show that when the signal-to-noise ratio is -16 dB,the detection probability of this method can reach 0.92 when the false alarm probability is 0.1.Meanwhile,amor wavelet is more suitable for non-stationary signal decomposition with better training and recognition ability.