Multiuser neural network detectors for large scale MIMO OFDM over Rayleigh Winner‐II channel

Abstract

Large scale (LS) MIMO orthogonal frequency division multiplexing (OFDM) technique satisfies the demands on performance and the service quality preferred in wireless communication system. Since numerous antenna terminals have been incorporated in the base station, multiuser detection is crucial for retrieving the data appropriately. Thus, the complexities of the detectors increase rapidly in large scale MIMO OFDM schemes. A unique neural network detection approach for multiuser detection in LS MIMO OFDM model is presented. The performance of multilayer perceptron (MLPNN) and radial basis (RBNN) detectors with minimum mean square (MMSE) MUDs are analyzed. The neural network multiuser detecting methodologies are also analyzed by the loss function in LS MIMO OFDM systems. The performance analyses of the detectors are also done by analyzing the radiation patterns of the antennas which is equipped at the both ends of the uplink channels. Radial basis MUDs have 23 dB bit error rate which gains over linear MMSE detector and of 14 dB than multilayer perceptron detectors specifically at 10 (À5) bit error rate. According to the bit error rate analysis of estimations at the output of the multiuser detector, the suggested neural network models are more effective in reducing multiuser interference in MU LS MIMO OFDM. Analysis also verifies, radial basis function (RBF) multiuser detector are exhibiting an optimal BER performance than multilayer perceptron detector with sufficient number of hidden neurons. This study thus emphasis a RBF MUD scheme for detecting multiuser signals in LS MIMO OFDM system due to its dynamic performance in low probability of bit error.