loading page

End-to-End Wireless Communication System Based on Deep Neural Network Channel Module
  • Ivan Ganchev ,
  • Zhanlin Ji
Ivan Ganchev
University of Limerick, University of Limerick

Corresponding Author:[email protected]

Author Profile
Zhanlin Ji
Author Profile

Abstract

The existing end-to-end (E2E) wireless communication systems require fewer communication modules, have a simple processing signal flow and do not require expertise compared to conventional wireless communication systems. However, in the absence of a differentiable channel model, it is impossible to train transmitters, used in such systems, which makes impossible to get optimal system performance. To solve this problem, an E2E wireless communication system with conditional generative adversarial networks (CGAN) for channel modeling has been proposed recently. Unfortunately, the training of CGAN is prone to instability, slow convergence, and inaccurate channel modeling, which affects the  system performance. To this end, we propose an E2E wireless communication system with a deep neural network (DNN) channel module as the unknown channel. Simulation results showed that the proposed channel modeling method has faster convergence, simpler network structure, and can reflect the behavior of real channels more accurately. In addition, the proposed E2E wireless communication system exhibits better bit error rate (BER) and block error rate (BLER) performance.