Research On Channel Estimation Based On Optimal Pilot Sequences In MIMO-OFDM Systems

Orthogonal Frequency Division Multiplexing(OFDM) is an improvedmulti-carrier modulation technique in wireless communication systems. It has highdata rate and can mitigate the effect of frequency-selective fading. Multiple InputMultiple Output(MIMO) can provide high capacities, good coverage and reliabletransmission which are needed in the fourth generation(4G) wireless communicationsystem by making the use of multiple antennas. The MIMO-OFDM system,combined by the two powerful technologies: MIMO antennas and OFDMmodulation, is the basic technique in the physical layer design of4G wireless system,and it can satisfy the increasing demand for high-performance. Considering thecomplexity and uncertainty of the transmission conditions in wirelesscommunication, channel estimation becomes a significant part of MIMO-OFDMsystems. The performance of estimation has a direct impact on the quality ofMIMO-OFDM systems.The channel estimation of the thesis is based on MIMO-OFDM systems. First,the characters of wireless channel and fundamental principles of MIMO-OFDMsystems are introduced. The focus here is a system model based on spatialmultiplexing and OFDM named VBLAST-OFDM.The process of channel estimation based on pilots is that first send a knownsignal at the transmitting end, then obtain the frequency response of the channel atthe receiving end by using an appropriate algorithm. There are three basic blocksaffecting the performance of channel estimation based on pilots. These are thechannel estimation algorithm, the pilot pattern and the signal detection part. In thisthesis, Least squares(LS) algorithm in time domain is chosen as the basic estimationalgorithm. The performance of LS algorithm has low complexity and is easy to berealized. The optimal pilot sequences are used as the pilot pattern. Analysis showsthat the estimation using optimal pilot sequences has good performance todistinguish subchannels, and it can reduce the complexity of LS estimationin someways. In VBLAST-OFDM systems, signal detection should be employed to differtransmission signals from different antennas at the receiving end. Zero-forcing(ZF)linear detection and maximum likelihood(ML) algorithm non-linear detection arestudied. As different methods have different performance, we can choose the properdetection method in a different case.In this thesis, improved algorithms are presented based on LS algorithm in timedomain. As the estimation accuracy of LS algorithm is not good, a method combinedadaptive filtering with LS estimation is proposed. After the initial LS estimation in time domain, we use an adaptive filter to make the estimation result more accurate.Normalized least-mean-square(NLMS) algorithm and recursive least-squares(RLS)algorithm are studied. The effect of different channel parameters on algorithms isalso considered. Theory analysis shows that the performances of MSE and trackingcapability in RLS algorithm are better than that in NLMS algorithm in the sameconditions. But the acquirement of the benefits is at the expense of computationcomplexity.