Signal Detection of OFDM Systems Based on Convolutional Neural Network

This paper proposes a convolutional neural network structure to improve the signal detection capability of an Orthogonal Frequency Division Multiplexing (OFDM) receiver. The network structure can replace channel estimation and signal detection modules of conventional OFDM receivers, improving performance while making the network structure simpler. The network structure mainly includes: input layer, two convolution-residual units, full connection layer and classification layer. In this paper, the random data generated through the 3GPP TR38.901 channel model is used to train the network model, and the trained model is finally applied to the OFDM system, which replace the channel estimation and signal detection module in the traditional OFDM receiver. Theoretical analysis and simulation results show that the method in this paper not only reduces the number of parameters, shortens the training hours and reduces the memory occupancy, but also achieves good SER performance.