Study On Channel Estimation In OFDM System

IntroductionThe target and requirement of wireless communication technology are totransmit data fast and reliably. Channel estimation is the basis of coherent detection,demodulation and cannel equalization. And it is also very important for OFDMtechnique to realize data communication at high speed. The receiver in OFDM systems can adopt incoherent detection or coherentdetection. The OFDM systems using incoherent detection do not need channelestimation. The estimation of transmitted symbols can be obtained by the methodof difference demodulation at receiver. It has the merits of non-requirement ofchannel state information and low complexity. The shortage is that it is onlysuitable for low data rate transmission system. Its performance is not as good asthat of OFDM systems based on coherent detection. In order to improve the datatransmission speed, the OFDM system exploits the receiver of coherent detectiongenerically. For the sake of resuming the data from the transmitter, the estimationof wireless channel is necessary at the receiver.The channel estimation methods can be divided into three categories. The firstone is the channel estimation method based on pilot or trained sequence. This kindof algorithm has good capability and is easy to realize. The transmition velocitydecreases because the introduced pilot symbols or trained sequence occupy someuseful bandwidth. The second class is blind channel estimation method based onthe limited characters of the transmitted information symbols and their statisticaltrait. This kind of method does not need pilot symbols or trained sequence, so thebandwidth is saved and it can improve the spectrum utilization efficiency. But ithas the disadvantages of slow convergence, high complexity and low capability offollowing the variation of wireless channel effectively. The third one is thesemi-blind channel estimation method by using the information from blind channelestimation algorithm and known sampling symbols to finish channel estimation. Itsolves the problems of spectrum waste from channel estimation based on pilotsymbols or trained sequence and high complication of blind channel estimationmethods. So the semi-blind channel estimation algorithm is regarded to be apromising way for channel estimation.This thesis mainly focuses on the channel estimation algorithm of OFDMsystems. In the paper the planar channel estimation algorithm based on pilot, theLS channel estimation algorithm based on trained sequence, the LMMSE channelestimation algorithm and blind channel estimation algorithm based on signalsubspace are analyzed. The semi-blind channel estimation algorithms based onsignal subspace, adaptive RLS and LMS algorithm are studied thoroughly. Aimingat the shortage of these methods, two algorithms of semi-blind channel estimationare proposed. One is the improved semi-blind channel estimation algorithm inOFDM systems based on signal subspace. While under color noise, in order toeffectively realize semi-blind channel estimation, the autocorrelation matrix ofreceived signal is computed by using gradient-based variable forgetting factor RLSalgorithm (GVFF-RLS). The other is the improved adaptive semi-blind channelestimation algorithm using gradient based variable forgetting factor RLS algorithm(GVFF-RLS) instead of RLS algorithm to approach the received noise signalsubspace, the convergence velocity of RLS algorithm is improved.1. An improved semi-blind channel estimation based on signal subspace inOFDM systemsThe principle of semi-blind channel estimation algorithm based on signalsubspace in OFDM systems is that auto-correlation matrices of received signal arecalculated with gradient-based variable forgetting factor RLS algorithm at receiver.Signal subspace and noise subspace are obtained by decomposing auto-correlationmatrices. The channel impulse response is estimated with the orthogonalitybetween noise signal subspace and channel impulse response matrices. But thismethod converges slowly in time varying channel and is only suitable for whitenoise.In order to overcome these deficiencies, this paper proposes an improvedsemi-blind estimation algorithm based on signal subspace. The main idea is thatwith gradient-based variable forgetting factor RLS algorithm, auto-correlationmatrices of received signal are calculated. Through simultaneous diagonalization ofthe overall covariance matrices of both received signal and noise signal, noisesignal subspace is estimated. So the channel impulse response and semi-blindestimation are obtained efficiently in color noise. The proposed method does notneed any signal models and any a priori assumption on the stochastic property ofthe noise signal. Computer simulation experiments show that the error performanceof the new method in color noise is as the same as in white noise and theconvergence rate is faster than the original method with invariable forgetting factor.The new method has good stability and fast tracking ability even when SNR is low.2. Improved adaptive semi-blind channel estimation in OFDM systemThe semi-blind channel estimation algorithm of signal subspace is easy torealize. But it is very complicated. In order to solve this problem, paper [46]proposed a method that it can obtain the noise signal subspace using adaptive filterinstead of signal subspace decomposition. The complication is reduced efficiently.The fundamental of adaptive semi-blind channel estimation based on RLSalgorithm is that noise signal subspace of received signal is attained by using RLSmethod instead of SVD algorithm. Channel impulse response is estimated with theorthogonality between noise signal subspace and channel impulse responsematrices. But this algorithm converges slowly in time varying channel. Aiming tothis problem, a new method is proposed by using gradient-based variable forgettingfactor RLS algorithm (GVFF-RLS) instead of RLS algorithm to approach thereceived noise signal subspace. It overcomes the disadvantage of slow convergenceexisting in most of semi-blind channel estimation algorithms and realizes theestimation in time varying channel. Simulations show that the convergence rate ofthe new method is faster than that of the original method with invariable forgettingfactor in multi-pathway fading time varying channel. In addition, it has goodstability and fast tracking ability when SNR is low.