Linear and Nonlinear Detections for Noncoherent Spatial-multiplexing Based MIMO-MASK

Multiple-input-multiple-output (MIMO) based on energy detection has the advantages of being insensitive to Doppler shift and low complexity. However, the square-law processing causes the nonlinear mutual interference of MIMO spatial-multiplexing signal detection. In this paper, linear and nonlinear detections are derived for spatial-multiplexing based MIMO multi-level amplitude-shift keying (MIMO-MASK) modulation. The linear low-complexity detection is based on the equivalent linear model with the complementary amplitude waveform and the energy difference operation at receiver; The other nonlinear detection using deep neural network is theoretically analyzed to be suitable for the nonlinear model of mutual interference. The former requires estimation of real channel information, while the latter does not require pre-estimation of channel information but requires supervised network training. The performances of both detection algorithms are verified by simulations, and both of them are insensitive to Doppler shift.