Wireless fading channel estimation method based on deep dense residual network

The invention discloses a wireless fading channel estimation method based on a deep dense residual network. According to the method, Add and Concatenate operations are introduced, a deep neural network DNN is improved by using a dense network DenseNets and a residual network ResNets, a deep dense network DeDNN and a deep residual network ReDNN are respectively constructed, the DeReNet is formed through series connection, gradient explosion and disappearance problems occurring in network training are inhibited, IQ signal features are automatically extracted through the neural network, and accurate estimation of a wireless communication channel is realized. According to the method, the accuracy and robustness of OFDM wireless communication channel estimation can be improved, and the generalization migration capability is high. 本发明公开了一种基于深度密集残差网络的无线衰落信道估计方法。该方法引入Add和Concatenate操作，运用密集网络DenseNets和残差网络ResNets改进深度神经网络DNN，分别构建深度密集网络DeDNN和深度残差网络ReDNN，通过串连组成DeReNet，抑制网络训练中出现的梯度爆炸和消失问题，通过神经网络自动提取IQ信号特征，实现无线通信信道准确估计。本发明