Research On Channel Estimation Technology In The High-speed Railway Environment By Used TDD LTE

With the rapid development of the information age, mobile communication not only has to meet users’ growing needs of high speed and high service bandwidth, but aslo needs to take into account the limited spectrum resources. This requires continuous research and ever improving technologies to promote the upgrade of the communication industry. Under this circumstance, there comes the LTE(Long Term Evolution) technology, which is currently the most important technical solution to meet mobile broadband users’ demand, as well as the future direction of mobile networks.The service quality of communitcation in high-speed railway will affect the evaluation of China Mobile Communications industry. In this case, the application of LTE-TDD technology in high-speed railway is promising and necessary.Due to the Doppler Effect, time-varying channel will bring ICI(Carrier Interference) in LTE-TDD systems. It is important to obtain accurate channel state information in MIMO-OFDM systems. Rician channel is a typical channel model, and Rician factor is an important parameter. So the research of channel estimation of time-varying Rician channel is valuable.In this thesis, we will focus on the Rician channel parameter estimation in the base of LTE-TDD protocol standard. We launched the following research:Firstly, we analyzed the high-speed railway channel model. We researched the characteristics of high-speed railway wireless communication; we studied the high-speed railway model proposed by 3GPP, along with WINNER II D2 A channel model; we studied Rician fading channel characteristics.Secondly, this thesis studied typical representative of the Rician factor estimation algorithm: the maximum likelihood estimation, moment estimateion and KS statistical estimation. Maximum likelihood estimation utilize the envelope and phase information of the received signal to estimate rician factor, which assures its best estimate performance; moment estimation relies on the envelope of the received signal to estimate the rician factor, commonly used is 1st, 2nd and 2nd, 4th order moments method; KS statistical estimation methods are based on the sampling point envelope mathematical statistics, the distribution function obtained will be compared with the known theoretical distribution function, according to some rules, to estimate the Rician factor. By using the simulation analysis method, we compare the performance and complexity of all types of algorithms when the LTE-TDD system is in high-speed railway situation. As a result, the second and fourth moments’ algorithm is estimated to achieve a good tradeoff between the performance and complexity.In this thesis, we studied the estimation algorithm in the time-vary Rician channel: least squares, linear minimum mean square error and shift scale least squares algorithm. In these three algorithms, the change of Rician factor does not affect the performance of least squares estimation algorithm; linear minimum mean square error and shift scale least squares algorithm will get the better performance with lagre Rician factor. The digital simulation proved the three methods in a tipical Rice fading channel’s mean square error performance, and also proved the one which suits long-term evolution uplink system better in the high-speed rail tunnel scene is shift scale least squares algorithm.Finally, we studied one of the steps, which is called the Fourier transform module, when the above algorithms were easily affected in Dightal Signal Processor chip realization. In this thesis, based on the core of TMS320C6670 quad-core multi-coprocessor DSP chip, we studied DSP’s multicore navigation system, then analysed and compared the performance of fast Fourier transform coprocessor emphasisly. Meanwhile, we analyzed the performance of dynamic and non-dynamic fast Fourier transform coprocessor adopted. By comparing the simulated data of MATLAB, the magnitude of calculation mean square error in verifying the fast Fourier transform is 10-8, the time complexity in handling 7×2048 point fast Fourier transform is 99.88 K clock cycle, the pricision error is 8.32×10-8, which basically meet the needs of experimental system.