Research On Channel Estimation Techniques In Mobile Broadband MIMO-OFDM Communication Systems

Large system capacity, high speed data rate and spectrum effciency are the tendencyof the modern wireless communications. The combination of multiple-input multiple-output(MIMO) and orthogonal frequency division multiplexing (OFDM) is one of the most promis-ing techniques for broadband wireless communications. For its advantage, MIMO-OFDMhas become a core technique in the high speed data down-link solution for the next gen-eration mobile wireless communication. Since both space-time decoding and data detec-tion require accurate channel parameters, channel estimation is one of the key techniquesfor MIMO-OFDM wireless communication systems. However, at the receiver antenna thereceived pilot signals are superposition of the pilot signals transmitted from different trans-mit antennas and experiencing the different independent channels respectively, which bringsmore diffculties for channel estimation in MIMO-OFDM systems.In general two classesof methods are available for channel estimation, one is based on training sequences, andthe other is blind channel estimation. Now the training sequences based channel estimationremains attractive in practice. This thesis mainly studies the correlative channel estimationtechniques based on training sequences in mobile broadband MIMO-OFDM communicationsystems.This thesis frstly studies the optimal training sequences for channel estimation inMIMO-OFDM systems. By exploiting the properties of co-cyclic Jacket matrices, this thesisproposes a novel scheme for obtaining the optimal pilot sequences. Based on the minimummean square error (MSE) of the least squares (LS) channel estimation, the optimal pilot se-quences distributed over one and multiple OFDM symbols and their necessary conditionsare derived respectively by constructing the co-cyclic Jacket matrices. This derived resultsare more generalized than those in references. It is shown that (1) The derived optimal pi-lot sequences are equi-powered, equi-spaced and phase-shift orthogonal;(2) The number ofpilots in the optimal pilot sequences is proportionable to that of the transmit antennas, itmultiples with increasing the number of transmit antennas. To improve the data effciency and reduce the PAPR, the optimal pilot sequences over multiple OFDM symbols are oftenused for channel estimation in MIMO-OFDM systems.In MIMO-OFDM systems, although the channel estimation scheme using the optimalpilot sequences over multiple OFDM symbols can achieve the perfect performance for slowlyfading channels, its performance deteriorates badly for very fast fading channels, which lim-its the use of the scheme in practical systems. To improve the performance of the optimalpilot sequences over multiple OFDM symbols in fast fading channels, this thesis proposes anovel enhanced channel estimation method using virtual pilot tones in MIMO-OFDM sys-tems. The virtual pilot tones only have the arithmetical meaning and don’t exist actually.If the virtual pilot tones are arithmetically superposed at the data locations over the specifcsub-carriers, then a special comb-type pilot structure is formed from the real and virtual pilottones. According to the characteristic of the optimal pilot sequences over one and multipleOFDM symbols, the virtual pilot tones are the same as the real pilot tones over the samesub-carriers in other OFDM symbols. Assuming that the superimposed virtual pilot tones atthe data locations over the specifc sub-carriers are transmitted from all transmit antennas,the corresponding virtual received pilot signals at the same locations are obtained from theneighboring real received pilot signals over the same sub-carriers by Wiener flter or linearinterpolation. Based on the LS channel estimation, the channel parameters can be obtainedfrom the combination of the virtual and real received pilot signals over each OFDM symbol.Simulation results show that the proposed channel estimation method greatly outperforms theprevious method for the optimal pilot sequences over multiple OFDM symbols in fast fadingchannels, as well as approaches the method for the comb-type optimal pilot sequences inperformance. Furthermore, the novel scheme requires much less pilot tones than comb-typeoptimal pilot sequences. Meanwhile, the virtual receive pilot tones only depend on one di-mensional Wiener flter or linear interpolation and has very low computational complexity.This thesis also proposes another enhanced channel estimation based on consecutive slid-ing windows and averaging, the scheme can also improve effciently the performance of theoptimal pilot sequences over multiple OFDM symbols in fast fading channels.This thesis also analyzes and evaluates the performance of the derived optimal pilotsequences in MIMO-OFDM systems, the results show that (1) The channel estimation us-ing the optimal pilot sequences can not only get the optimal performance, but also havevery low computational complexity in MIMO-OFDM systems;(2) The derived optimal pilotsequences have very high peak-to-average power ration(PAPR), which results in the highPAPR for the OFDM symbols that include the optimal pilot sequences;(3) Increasing the number of the OFDM symbols where the optimal pilot sequences distributed, the PAPR be-come lower. Taking into account practical constraints, the optimal pilot sequences with lowPAPR must be used. To obtain the optimal pilot sequences with low PAPR, the pilot se-quences are weighted by random phase vectors according to the idea of the Partial TransportSequences(PTS). For simple implementation in practice, the PN sequences are selected asthe random phase vectors, the optimal pilot sequences with low PAPR are obtained. Thesimulation results show that the PAPR of the enhanced optimal pilot sequences is reducedgreatly, and the optimal performance remains.The results also show that the PAPR of theoptimal pilot sequences over multiple OFDM symbols is lower than those over one OFDMsymbol.