Channel Estimation Of Ofdm System Based On Bem Fast Time-Varying Channel Model

As one of the multi-carrier transmission technologies, Orthogonal Frequency Division Multiplexing (OFDM) can provide high data rate, high spectrum efficiency and strong ability of combating multipath interference. As the high speed railway in our country is booming, the extreme high velocity of mobile terminals makes the wireless channel change quickly, thus the wireless communication system faces new challenges. In OFDM system, real-time and accurate channel estimation is especially with fast time varying channel, real-time and accurate channel estimation is the basis of coherent detection, demodulation and equalization. Channel estimation is the key factor in maintaining the performance of the whole system. This thesis pays attention to pilot-based channel estimation technology in OFDM system with fast time-varying channel.Firstly the characteristics of fast time varying channel are discussed, including Doppler Effect, multipath effect and the fading properties. Based on these properties, together with time-selectivity and frequency selectivity of fast time-varying channel in OFDM system, the thesis studies basis expansion model (BEM). Some simulations are done to test several widely used BEM models. Then, the principle and basic structure of general OFDM system is given.The thesis then discusses different insertions of frequency domain pilots and comb-type pilot aided channel estimation. Meanwhile, BEM channel model is widely used because it can help reduce the estimation complexity. Using CE-BEM channel model, two channel estimators, Least Square (LS) estimator and (Linear Minimum Mean Square Error) LMMSE estimator are analyzed and compared. To improve the accuracy of channel estimation based on CE-BEM model, we consider the following two methods:(1) use DPS-BEM channel model in LMMSE estimator to improve estimation accuracy with sacrifice of computation complexity;(2) change BEM model order adaptively with the doppler frequency to get small estimation error under the condition of high doppler spread. Facing the high computation complexity of LMMSE estimator, the following parts try to discuss a channel estimator based on Legendre-BEM model, this method has lower complexity compared with LMMSE estimator. In this method, we use received pilot signal to calculate the Fourier coefficients of the channel taps, then these obtained Fourier coefficients are mapped to the Legendre space to get the coefficients of Legendre-BEM. The operation of matrix inversion is no longer used in this new method thus reduces estimation complexity. Otherwise, Legendre-BEM model help reduce the Gibbs effect in CE-BEM model. The thesis gives details of this new method and LMMSE estimator, the simulation results of this two methods are displayed.Finally, the comb-type pilot frequency domain is studied. The widely used FDKD structure is modified and a new clustered comb type pilot structure is proposed. The simulation results show that this new comb-type pilot can help reduce the estimation error. Considering the spectrum efficiency, we simplify the guard pilot in each comb-type pilot cluster according to the analysis of inter-carrier interference. This simplified comb-type pilot structure reduces several guard pilots, at the same time ensures the estimation performance.