A Study Of Signal Detection In MIMO Systems Over Flat-fading Channels

Multiple-input multiple-output (MIMO) technique significantly improves transmission efficiency by exploiting spatial sources. In MIMO systems, multidimensional signal detection with high complexity is involved, which results in obstacles for implementation. On the purpose of pursuing good tradeoff between performance and complexity, the coherent and noncoherent detection are studied respectively.V-BLAST (Vertical-Bell Labs lAyerd Space-Time) systems which belong to spatial multiplexing architectures have to adopt coherent detection. Sphere decoding (SD) is an efficient algorithm to implement maximum-likelihood (ML) detection, and the strategy of initial radius selection and assignment has significant impact on the performance and complexity. By utilizing the statistical property of a V-BLAST signal model and applying the idea of statistical pruning, strategies of linear radius assignment and further a quasi-linear assignment based on performance analysis are proposed. The strategies alleviate some defects of the conventional sphere decoding, such as high complexity at low signal-to-noise ratios (SNRs) and average complexity sensitive to SNRs.The automatic sphere decoding (ASD) proposed recently is the most efficient among the existing sphere decoding algorithms. But it requires a large number of comparisons. Considering that comparisons have nontrivial impact on time complexity of algorithms in the present DSPs technology, a pruning ASD (PASD) is proposed on the premiss of keeping the efficiency of ASD. With a proper configuration, a good tradeoff is achieved between the performance and complexity. For high order modulations, PASD is able to reduce the number of comparisons by more than 30% as compared to ASD without much penalty of performance.Differential unitary space-time modulation (DUSTM) in conjunction with noncoherent detection avoids MIMO channel estimation and is suitable for fast fading channels. Under the assumption of continuous fading, a multiple symbol differential ASD is developed by embedding a recursion of an ML metric in ASD. The algorithm accomplishes ML detection with a complexity square in the length of observation window, and provides a framework to the design of multiple symbol differential detection (MSDD) based on SD. Furthermore, by ignoring nonsignificant factors in the performance, a multiple symbol differential approximate ASD with reduced computational complexity and a multiple symbol differential pruning ASD with reduced the number of comparisons are presented. The proposed algorithms are suitable for arbitary unitary space-time constellations, and eliminate the error floor in fast fading channels caused by MSDD with quasi-static channel assumption. Compared with the existing typical algorithms with continuous fading assumption, i.e. multiple symbol decision-feedback detection (MS-DFD) and noncoherent sequence detection (NSD), the performance of the proposed algorithms are much superior to that of MS-DFD and a little inferior to that of NSD, while the complexity is lower than that of MS-DFD in most cases and significantly lower than that of NSD.For two coded DUSTM systems, i.e. BICM (Bit-Interleaved Coded Modulation) -DUSTM and Turbo-DUSTM systems, soft-output and soft-input soft-output detection based on MSDD are derived respectively with quasi-static channel and continuouse fading assumption. In these algorithms, the MSDD algorithms for uncoded DUSTM systems are used to simplify the computation of a posteriori probability. Compared with most of the existing soft detectors based on single symbol differential detection (SSDD), the proposed algorithms take good advantage of the correlation of fading channels, and have stronger abilities to accommodate fast fading mobile environments.In the schemes of extending MIMO systems to wideband applications, MIMO-OFDM (Orthogonal Frequency Division Multiplexing) systems decompose a frequency-selective channel into a set of parallel flat-fading subchannels, which makes the detectors for flat fading MIMO channels applicable to the systems directly. For V-BLAST-OFDM systems, by utilizing the correlation among subchannel frequency responses and combining the idea of group ordering with interpolation-based QR decomposition, an ordering interpolation-based QR decomposition algorithm for two transmit antennas is presented. The algorithm obtains the QR decomposition of all the subchannel matrices with different ordering directly from several known QR decomposition of subchannel matrices, which can be used as a preprocessing method with low complexity for V-BLAST detection on each subchannel. For a system in which differential space-frequency modulation is accomplished within one MIMO-OFDM symbol, MSDAASD and its soft counterpart are applied to the system with a little modification. Compared with SSDD in the literature, the algorithms enhance the flexibility to multipath spread.