Neural Network-Based Receiver in MIMO-OFDM System for Multiuser Detection in UWA Communication

High quality of service and high transmission rate is the demand for future underwater acoustic (UWA) communication which is achieved through the implementation of multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) system in the UWA communication. However, the quality of the MIMO-OFDM system has faced multiaccess interference (MAI) at the receiver due to the interference from co-channel users. Therefore, multiuser detection (MUD) technique is needed at the receiver of the MIMO-OFDM system to suppress the effect of MAI. The novelty in this research is that MUD is achieved using multilayer perceptron (MLP)-based neural network (NN) detector at the receiver of the MIMO-OFDM system in the UWA communication. The MLPNN detector achieved the MUD at the receiver of the MIMO-OFDM system through the adaptation of NN weights and bias weights in the backpropagation (BP) algorithm. The transceiver model of the MIMO-OFDM system in underwater is implemented using BELLHOP simulation system. The bit error rate (BER) performance of the MLPNN detector towards MUD is analysed and is compared with that of existing detectors (matched filter (MF) detector, decorrelating detector (DD), minimum mean square error (MMSE) detector, and multistage conventional parallel interference cancellation (PIC) detector) in the UWA network. Proposed MLPNN detector outperforms in BER analysis over existing detectors in the UWA network.