Signal Detection in Uplink Pilot-Assisted Multi-User MIMO Systems with Deep Learning
Hang Hua, Xiaoming Wang, Youyun Xu · 2019
In this paper, we present a deep learning model for signal detection in uplink pilot-assisted multi-user MIMO systems and design two system frameworks for two cases of channel information at receivers, respectively. First, we consider the case that the channel matrix is known at base station (BS), the deep neural network (DNN) model can be seen as a detector and the transmission symbols are detected through channel matrix and received signal. Next, we consider the case in which the channel matrix is unknown at BS. The pilot-assisted DNN model we design can be regarded as a combination of channel estimator and signal detector. In this pilot-assisted DNN model, the channel information is indirectly reflected by pilot signal and the transmission symbols can be detected directly. The training data used for DNN model training is generated by simulation. When training stage is completed, the trained DNN model can be deployed for signal detection. The simulation results shown that the DNN-based detector with perfect channel state information can get better performance than conventional minimum mean squared error (MMSE) detector. Moreover, when channel matrix is unknown at BS, the pilot-assisted DNN-based detector can recover transmitted signal directly in an end-to-end manner without explicitly estimating channel and can get better performance than MMSE detector with least square (LS) channel estimation.