Research On Massive MIMO Detection Technology Based On Maching Learning

Massive Multiple-Input Multiple-Output(Massive MIMO)technology is one of the core technologies of 5G,which can greatly increase the spectral efficiency.Due to the large number of antennas,the computational complexity of detection increases significantly,posing a huge challenge to traditional detection algorithms.The application of deep learning for MIMO signal detection can achieve high computational parallelism and it is an important technical approach to solve high-dimensional signal detection problems.In the existing research,DetNet is a typical representative signal detection network of deep learning,with high performance.However,we found that DetNet still requires improvement.This paper mainly improves the DetNet network in two aspects:(1)simplification of the detection network.(2)extended detection model to high-order modulation.The simplification of the network includes three aspects:network input,network connection structure and loss function of the network.Then,we extend the improved network to Massive MIMO detection of high-order modulation.In high-order modulation,each modulation symbol is mapped into a high-dimensional binary vector as a detection output of the network,and then the vector is linearly weighted to realize demodulation of high-order modulation symbols.The simulation results show that this detection method is suitable for the scenarios with more antennas,and the detection performance can be improved with the increase of the number of transmitting and receiving antennas.