Research On Signal Detection Techniques For MIMO Systems

With the increased number of mobile users and the rapidly growth of broadband multimedia services, higher transmission rates and better performance are highly demanded in wireless communications. Researches on techniques for future mobile communication systems have been carried out. Among the numerous solutions, Multiple Input Multiple Output (MIMO) is regarded as one of the key techniques for next generation communication systems. MIMO technique can greatly improve the system capacity and performance without increasing power and bandwidth, which is high in favor due to the limitation of spectrum resources today. Vertical Bell Laboratories Layered Space-Time (V-BLAST) proposed by Wolinansky and Foschini is a typical application of MIMO technique and its spectrum efficiency is much higher than that of the traditional communication systems. Because of the low diversity order, the performance of traditional V-BLAST detection algorithms is not satisfying. Supported by the National High Technology Research and Development Program of China under Grant No. 2001AA123014 and National Science Foundation of China under Grant No. 60496310, this dissertation makes researches on signal detection techniques for MIMO systems over independent flat fading channels, aiming at improving the receive diversity order, ameliorating the Bit Error Rate (BER) performance, and decreasing the computational complexity, without loss in spectrum efficiency.The principle of amicable orthogonal designs helps to realize maximum likelihood (ML) detection with linear complexity, but the spectrum efficiency of this scheme is low. Relaxing the constraints in amicable orthogonal designs, we propose the concept of "quasi-amicable orthogonal designs". With the combination of this concept and the idea of layered space in V-BLAST, a MIMO transceiver scheme using grouped quasi-amicable orthogonal designs (GQOD) is proposed, which can achieve the same spectrum efficiency with V-BLAST system. In a MIMO system which allows multiple sub-streams to be transmitted simultaneously, the sub-streams are divided into several groups and symbols in each group are mapped to the corresponding transmit antennas with sets of quasi-amicable orthogonal vectors. It should be mentioned that there is no restriction among the sets used in different groups. At the receiving end, inter-group interference is cancelled first. For symbols in the same group, the joint detection of the symbol vector is transferred to the detection of the separate symbols with quasi-amicable orthogonal principle, and thus linear detection is realized. Theoretical analysis and simulation results show that GQOD transceiver scheme can decrease BER dramaticly, and the computational complexity of this scheme is lower than that of the present V-BLAST detection algorithms.Although GQOD transceiver scheme provides some performance gains, the diversity order of this scheme is still lower than the number of receive antennas. In order to improve the BER performance further, a new MIMO transceiver scheme using singular linear transformation (SLT) is proposed. Firstly, the transmitter determines the proper time interval d according to the coherence time of the wireless channels. Then several signal vectors with an interval of d are combined together by a singular matrix, and the result vectors are transmitted at the corresponding time slots. The receiver designs zero forcing (ZF) filter according to the singular matrix used in the transimtter. Then the received signal vectors at different time slots are fed into the ZF filter to complete the coarse estimation of the transmitted signal. Basing on the coarse estimation, the receiver accomplishes the fine estimation using parallel interference cancellation. If the time interval d is large enough to maintain the incoherence among the channel matrices at the corresponding time slots, the ZF filter can achieve a high diversity order. Although the procedure of fine estimation can not provide further diversity gain, it decreases BER effectively. Simulation results show that the BER curves of SLT transceiver scheme are nearly parallel to those of the ML algorithm, which implies that the two signal detection methods almost have the same diversity order. Compared with the V-BLAST algorithm using ordered successive interference cancellation, SLT transceiver scheme provides dramatic performance improvement with a litter higher complexity.SLT transceiver scheme provides high diversity order, but there is 4~5 dB performance loss compared with the ML algorithm. With altering the search method of ML algorithm and adopting the idea of search radius in sphere decoding, a new V-BLAST detection algorithm using subset search (SS) is proposed, which can achieve ML or near-ML performance. With a certain constraint, the receiver searches through the full signal space and obtains a subset. Then the final decision is made by selecting the proper signal vector from the subset with ML principle. Since the search radius in the constraint influences the computational complexity greatly, two branch algorithms are proposed to calculate the search radius: the SS V-BLAST based on ZF detection and the SS V-BLAST based on variance of noise. The difference between these two algorithms is that the former has no use for prior information of noise, while the latter needs the variance of noise. Furthermore, a variable betweent 0 and 1 is introduced to realize a proper balance between BER and computational complexity. That is, the variable helps to decrease the complexity greatly with a cost of small loss in BER performance. Simulation results show that SS V-BLAST detection algorithm can achieve ML or near-ML performance with low complexity.Three solutions, GQOD transceiver scheme, SLT transceiver scheme, and SS V-BLAST detection algorithm, are proposed to complete signal detection for MIMO systems over independent flat fading channels. These three solutions not only achieve the same spectrum efficiency with V-BLAST system, but also outperform the present V-BLAST algorithms in consideration of BER performance and computational complexity. The proposed solutions have their own advantages and disadvantages. SS V-BLAST algorithm can achieve the best BER performance but with the highest complexity. GQOD transceiver scheme has the lowest complexity while its BER performance is worse than the other two solutions.