<FONT face=Verdana>In orthogonal frequency division modulation (OFDM)

channel′s variation introduces intercarrier interference

making the accurate estimation of time-varying channel more complicated. For the number of channel parametres to be estimated at the receiver is greater than that of the received signal

the equation of channel estimation has no definite solution. To solve this problem

Basis Expansion Models (BEM) are usually used to approximate the channel′s variation. The problem of channel estimation is reduced to estimate the expansion coefficients

making the number of parameters to be estimated significantly reduced. The performance of widely used BEMs are analyzed. Based on oversampling and baseline tilting

an improved BEM is proposed to reduce the effects of Gibbs phenomenon and therefore increase the expansion accuracy. The performance of the proposed BEM is independent from the channel statistics and can be used for time-varying channel estimation in OFDM systems.