Research On The Dectection And Equalization Techniques In Multicarrier And Multiantennas Systems For The Future Communications Syetems

Rapid progress of mobile communication and internet networks in recent years has seen a dramatically growing demand for broadband wireless access. Multicarrier technique has such advantages: having high frequency spectrum efficiency; being capable to mitigating multipath effects and enabling easy equalizer; easy implementation by applying fast Fourier transform (FFT) and dynamic adaptation to channel characteristics. Thus it draws more and more research interests. While MIMO technique can substantially improve spectrum efficiency and system reliability, it also becomes the focus of researchers. In this paper we do research on detection and equalization techniques in multicarrier and multiantenna systems.We first discuss common physical models of transmission channels based on their basic characteristics. Then we investigate basic problems about how to model the typical channels. Further we introduce multicarrier and multiantenna systems' basic principles, including OFDM basics and system parameters configuration; informatics about multiantenna systems, diversity technique, linear space-time coding and its detection, space-time multiplexing and so on.With high carrier frequency in future communication systems and long OFDM symbol caused by and large subcarrier number, high mobile velocity that will result in fast-fading channels, will make the OFDM system suffered by the intercarrier interference (ICI). Hence we first propose a globally optimized random search algorithm and prove its asymptotic global convergence; then to improve its convergence velocity and reduce implementation complexity, we propose a practical sub-optimal random search algorithm and analyze its implementation complexity. Simulation results indicate that compared to the existing methods based on the classical zero-force detection and time domain parallel interference cancellation, the proposed sub-optimal algorithm can greatly reduce BER of the detection.The assumption that makes it feasible to detect the orthogonal space frequency code OFDM system by the classic linear maximum likelihood detection is that channel's frequency response keeps constant in OFDM subcarrier group taken up by space-frequency coding. Because the limitation of the number of subcarriers and the nonideality of the real transmission channels, channel's frequency response could not keep constant in real systems. Classical maximum likelihood detection and parallel interference cancellation will give rise to severe BER in real systems because multipath channels are rather rigorous or the number of subcarriers is few. Thus we propose two serial iterative linear minimum mean squared error (LMMSE) estimation detection methods. In the first LMMSE method, we update detection symbol's first order and second order statistics by iteration; to reduce its complexity, we propose a second LMMSE method using decision feedback. Simulation results indicate, compared to classical maximum likelihood detection and parallel interference cancellation method, the two proposed methods can greatly reduce detection BER and obtain close-to-ML performance. The proposed methods can be generalized to detect the orthogonal space-time coded OFDM system in time-varying multipath channels.We also discuss the soft equalization for the block transmission in single carrier MIMO systems over frequency-selective channels. Though BCJR and Viterbi algorithm can obtain optimal detection, the complexity improves exponentially with outputs number and channel memory length, and thus they are hard to implement. For full-rank MIMO channels (the receiver has more antennas than the transmitter), we propose two low complexity sub-optimal soft equalization algorithms with quasi-ML performance. Both algorithms apply sliding window based sub-signal model. In the first algorithm, we use the constant modulo property of the transmit symbols, and we propose a symbol metric based max-log-MAP sphere decoder (MLMSD). This MLMSD has three characteristics: it is capable to use the priori information of the transmit symbols updated by the nearby sub-models; it uses the LMMSE filter based pre-processing; it uses an improved greedy enumeration strategy. For non-PSK constellation, to use the first algorithm, we must transform the QAM signaling to the QPSK one first. The first algorithm has high complexity when receiver diversity is low, or the rank-deficiency of the sub-signal model is raised after the QAM to QPSK transform. To combat the disadvantages of the first soft equalizer, we propose the second soft equalization method. This method combines soft interference cancellation (SIC) and the existing MLMSD. First it uses SIC to improve sub-signal model and obtain a nearly equivalent full-rank sub-signal model. Then it applies the MLMSD in this model, gets transmitted bits posteriori information. And according to the posteriori information, it updates first and second order statistics, which will be used by the SIC operations later. Simulation results indicate that for the PSK constellation, the first soft equalizer can get close-to-ML performance with lower complexity compared to the probabilistic data association (PDA) filter based soft equalizer and the second soft equalizer; the second soft equalizer's complexity grows a little with the decreasing of the receiver diversity but is lower than that of PDA filter based soft equalizer. And in this case it still has a close-to-ML performance, which is different from the PDA filter based soft equalizer and the first equalizer because their performance will have increasing obvious gap with the ML detection.