基于深度神经网络的无线衰落信道估计模型

DeReNet: WIRELESS FADING CHANNEL ESTIMATION MODEL BASED ON DEEP NEURAL NETWORK

摘要: 为提高 OFDM (Orthogonal Frequency Division Multiplexing) 系统通信衰落信道估计性能，提出一种基于深度神经网络的信道估计模型 DeReNet。通过串联深度密集网络和深度残差网络，抑制网络训练中出现的梯度爆炸和消失问题。将该模型与 LS (Least Square)、FC-DNN (Full Connection Dense Neural Network) 和 SimNet (Simplified Deep Neural Networks) 模型进行仿真实验对比，结果表明，在莱斯衰落环境下该模型的信道估计性能更好，能有效提高信道衰落的估计精度。

Abstract: In order to improve the communication fading channel estimation performance of OFDM (Orthogonal Frequency Division Multiplexing) system, a channel estimation model DeReNet based on deep neural network is proposed. DeReNet suppressed the gradient explosion and disappearance problems in network training by cascading deep dense networks and deep residual networks. In order to verify the effectiveness of DeReNet, DeReNet model was compared with LS, FC-DNN and SimNet models. The simulation results show that in the Rice fading environment, the channel estimation performance of DeReNet is better than that of three compared models, and the DeReNet model can effectively improve the estimation accuracy of channel fading.