Performance of Deep Learning Methods in DF Based Cooperative Communication Systems

Abstract

In recent years, deep learning (DL) based communication systems have gained great importance. In this paper, channel estimation for Rayleigh fast fading channels is proposed by applying three different deep learning algorithms which are multilayer perceptron (MLP), convolutional neural network (CNN), and long-short term memory (LSTM). Firstly, the maximum likelihood (ML) estimation is implemented by employing the Monte Carlo simulation program. Training and test samples for models are generated by combining the complex data consisting of real and virtual parts obtained at the channel output. For training process, channel output is used for the input of DL models whereas the channel coefficient matrix is used for the output of the models. DL algorithms have shown good performance comparable to ML estimation in terms of bit error rate (BER). From the simulation results, DL algorithms can be seen a promising method for channel estimation in cooperative communication.