Study On Blind Space-time Equalization In MIMO System Based On ICA

1,IntroductionMultiple-Input Multiple-Output (MIMO) techniques, as a major focus of research topic in wireless mobile communications, have aroused more and more attention in resent years. Important improvement in throughput and spectrum utilization can be achieved without additional bandwidth when multiple antennas are applied on both the transmitter and receiver side.MIMO wireless communication systems usually work in complicated fading environments. In such scenarios, co-channel interference (CCI) caused by other users adds to the intersymbol interference (ISI) generated by multi-path propagation, hindering the reception of the signals. Additionally, noise and interference induced by signal processing also result in signal distortion. Both channel identification and data detection are more challenging in MIMO communication system. Especially, its ability of space–time equalization will affect even determine the transfer capacity and spectrum utilization of the wireless channels directly.Signal processing techniques for space-time equalization aim at the cancellation of CCI and ISI at the receiver antenna output, and the recovery of the transmitted users'data. At present, three classes of channel equalization schemes are suggested: equalization based on the training sequences, blind equalization and semi-blind equalization. In MIMO systems, blind space-time equalization scheme is praised because their complicated transmitting environments. Instead of relying on the training sequences, it makes use of the statistical properties of the output signals. Generally, constant modulus, finite alphabet, cyclostationarity or the non-Gaussian distribution of the digital communication signals can be exploited. The following blind methods are often used: Constant Modulus Algorithm (CMA), Maximum Likelihood Estimation (MLE), sub-space decomposition method, and so on. All of them can equalize the frequency-selective finite impulse response (FIR) channel.2,Linear MMSE Equalization Based on Channel IdentificationIn space-time processing, the Maximum Likelihood Sequence Estimation(MLSE) is the optimal criterion, but its computational load can be prohibitive in scenarios involving a large number of users and highly dispersive channels.Trading off complexity for performance, linear receivers are based on the estimation of a linear transformation fulfilling certain sub-optimal criterion, such as Minimum Mean Square Error (MMSE) detector. Combining the advantages of time- and space-only processing, and considering the contradiction between the cancellation of CCI-ISI and the enhancement of noise, MMSE detector has an outstanding performance in scenarios with low SNR.This thesis mainly focuses on the blind space-time equalization of multi-user MIMO system. Blind multi-channel equalization can be performed with or without previous channel identification. Channel identification-based equalization presents the main drawback that inaccuracies in the channel estimate have a detrimental effect on the signal detection stage. However, a two-stage processing is useful for blind space-time equalization. In the first stage, second-order statistics (SOS) are able to cancel ISI by taking advantage of the structural properties of the channel and the source data matrices (time equalization). An instantaneous linear mixture of the source signals (i.e., a CCI-only cancellation problem) is resolved in the second stage. It can then be solved using source separation techniques based on properties inherent to digital signals (space equalization). Furthermore, knowledge of the channel makes it possible to select the equalization delay which yield considerable performance improvements with significant computational savings. Tong's method for a SIMO model is extended to MIMO system, to realize the blind channel identification based on the decomposition of the autocorrelation matrices of the observed vector at tow different time lags. Then the users'signals are detected by the conventional MMSE equalizer.3,Traditional MMSE Equalization Refined by ICAIn conventional MMSE equalization, the autocorrelation matrices of the received signal are estimated by the samples of the observed vector, i.e. exploiting time average instead of statistical average. Imprecision due to finite sample size, in the estimation of the channel matrix and the sensor covariance matrix has a negative impact on the detection of the transmitted data symbols.Alternatively, the mutual statistical independence between the users'signals can be exploited through the use of Independent Component Analysis (ICA) basedon higher-order statistics (HOS). ICA is proved useful in refining conventional linear detections without any prior parameters. The rationale behind MMSE-ICA detection consists of taking advantage of the available channel estimate as an initial point in the ICA search. Two main benefits can be derived from this refinement. Firstly, since conventional detections only make use of SOS, the exploitation of HOS by ICA is expected to mitigate performance drops caused by estimation errors at the channel identification stage, and to improve robustness of linear detections. Secondly, if these channel identification errors are moderate, ISI-CCI suppression is implicitly carried out during MMSE equalization, and the initialization provided by the channel estimate may already be quite close to the ICA solution, thus decreasing the convergence time and computational complexity of the ICA post-processing stage.The purpose of blind equalization in multi-user MIMO system is to demodulate all the users'data simultaneously. The previous section has illuminated how all time-shifted versions of each user are recovered by the ICA refined detector. However, most of the detected signals are redundant; a single time delay suffices for each user in practice. A simplified MMSE-ICA detection scheme was originally proposed for fixed-delay equalization, which leads to a consequent decrease in computational complexity by extracting only one time-shifted version for each user. Whereas, the choice of the fixed delay is somewhat arbitrary and may indeed be statistically suboptimal. Herein, we improve on the original definition by allowing arbitrary delays for respective users. Hence, from the choice of the equalization delay providing the best MMSE performance for each user, it is possible to compute the equalizer that will detect each source signal with the lowest MSE. This simplified scheme is not only more computationally efficient, but also outperforms the full MMES-ICA detector, as has been illustrated in section 4.4,Simulation and ConclusionWe simulate a wireless communication system composed of M = 5 simultaneous QPSK-modulated users across a frequency-selective block fading channel introducing ISI from a maximum of K = 4 consecutive symbol periods. The channel filter taps are randomly drawn from a complex Gaussian distribution and hence model a Rayleigh propagation environment. A smoothing factor of L = 5 combined with a spatio-temporal diversity level of N = 10. The zero-meanadditive white Gaussian noise at the sensor output has covariance R n =σ2I N is independent of the data sources. The simulations of the ICA refining algorithm using MATLAB are performed. As is shown in the experiments, the performance gains are achieved at only a modest increase in computational load relative to the conventional receiver. Furthermore, the optimal delay based equalization has exhibited improved performance and lower computational cost in moderate to high SNR and sample size conditions (These conditions can be considered as realistic in practical scenarios.).