Two-Stage LMMSE/DNN Receiver for High-Order Modulation

Abstract

High-order modulation has been seen as a promising technology to increase the spectral efficiency of wireless communications. Unfortunately, higher-order signals are susceptible to power amplifier (PA) nonlinearities and multi-path effects of the channel, leading to nonlinear distortion and severe inter-symbol interference. To guarantee the reliable detection of high-order signals, a two-stage receiver consisting of linear minimize mean squared error (LMMSE) equalizer and deep neural network (DNN) demodulator, namely TSLD receiver, is proposed, where LMMSE equalizer and DNN are deployed to eliminate inter-symbol interference and handle the amplifier nonlinearity, respectively. To facilitate the implementation of our proposed receiver, the specific pilot structure is designed, where low-order modulations are used for LMMSE equalizer and high-order modulations are used for DNN training. Simulation results show that our proposed TSLD receiver for high-order demodulation could recover the transmitted symbols effectively, even under serious multi-path and nonlinerity scenarios.