Signal Detection Research For MlMO-OFDM Systems

Comparing with other types of wireless systems, MIMO-OFDM system is ableto provide higher data transmission rate, spectral efficiency and system capacity. It hasbecome a focus in wireless communication research and been employed by the latestverson of many protocols and standards for attaining higher quality of wirelesscommunication.This paper takes a research on the signal detection algorithm of MIMO-OFDMsystems receiver, which as a key part, directly influence the performance of the wholecommunication system. Through analysising the mathematical principle ofconventional detection algorithm and sub-optimal ML detection of MIMO-OFDM,simulating those algorithm in Matlab, this paper summerizes the weakness of thosealgorithm and points out the improvments.Neural networks with Hopfield structure is able to transform the process ofminimizing the objective function into the process of network energy functionmonotonic decreasing. This paper considers the ML detection as the objectivefunction, by tranforming it into complex valued neural network energy function, thevalue of energy function is decreasing while the neural network feedbacks. Thisequals to the value of maximun likelihood decreasing. The mathematical derivation ofthe Hopfield network parameters are given, the BER performance of the proposedalgorithm is evaluated via Matlab simulations and compared with others conventionaldetection algorithms. It is shown that the proposed algorithm is a sub-optimal schemewith lower computations for MIMO-OFDM signal detection with BPSK modulation.Seletive Spanning with Fast Enumeration (SSFE) algorithm is a sub-optimal MLdetection specifically for MIMO-OFDM systems with high-order modulation scheme.But the Fast Enumeration does not fit for low-order modulation scheme well, in whichboundary constellation points dominate. Because it calculates some fake constellationpoints by fixed fomula, those are not belong to constellation map at all. This paperproposed the Direct Enumeration to avoid producing fake constellation points throughFast Enumeration. So the SSDE detection algorithm is able to apply in MIMO-OFDMsystems with QPSK or16QAM modulation scheme. The simulation results andanalyses show that the SSDE is able to give a closed BER performance to MLDetection with extremely low computational complexity comparing with MLdetection.