Channel Estimation For OFDM Systems Over Doubly Selective Channels

Orthogonal frequency division multiplexing (OFDM) is one of the importanttechniques in future wireless communications to achieve high quality, high data rate,high spectral efficiency, high bandwidth, due to its robustness against frequencyselective fading, and realization reliable transmission of multimedia services，such asvoice, data, video，et al. In the future broadband mobile communications, channelestimation becomes more challenging for OFDM systems in complicated timefrequency doubly selective fading channels，so the research of channel estimation isvery valuable for OFDM systems over doubly selective fading channels.This paper studies pilot-based channel estimation for OFDM system over doublyselective fading channels.Firstly，the paper discusses several basis expansion model（BEM），which arecommonly used for modeling doubly selective fading channels. On this basis, Legendrebasis expansion model（L-BEM）is chosen and used to approximate time-varyingchannel, two based on L-BEM channel estimators are designed，they are the least square（LS）estimator and the linear minimum mean square error（LMMSE）estimator. Inorder to reduce model errors，the accurate mathematical model for channel estimation isestablished by using the structure characteristic of frequency domain channel responsematrix. What’s more, the parameters are reasonable adjusted to suppress the influence ofdoppler shift to channel estimation performance.Secondly, In terms of the deficiencies of channel estimation algorithm based on thecomplex exponential basis expansion model (CE-BEM) in OFDM systems underdoubly selective fading channels, this paper proposes channel estimation algorithmbased on windowed basis expansion model. The algorithm can be applied to reducespectrum leakage which exists in CE-BEM through the time domain windowingprocessing, and suppress the influence of doppler shift to estimation performance; andthen the estimated coefficients of CE-BEM are converted by designed orthogonalprojection, and further improve the channel estimation accuracy. On the basis of thisstudy, another new method has been proposed. The estimated coefficients are convertedby the inverse reconstruction method instead of the orthogonal. This proposed methodstill effectively improve the channel estimation performance.Finally, this paper propose a new based on channel correlation pilot-aided algorithm for channel estimation in the frequency domain，by exploiting the statistical models oftime domain channel response matrix and frequency domain channel response matrix,and combining the structure of frequency domain channel response matrix. Thisalgorithm reduces model errors through making full use of channel correlation andconsidering the out-of-band interference from frequency domain channel responsematrix, and further suppress the influence of ICI to channel estimation performance byadjusting the bandwidth and the length of pilot cluster.Experimental simulation results demonstrate that all of above-mentioned proposedalgorithms can effectively improve the channel estimation performance.