Low Complexity Channel Estimation for Minimizing Edge Effects in OFDM Systems

Abstract

Channel estimation is critical to receiver performance in an OFDM system. Pilot symbols are typically interspersed with data symbols among the sub-carriers to aid channel estimation. Standard algorithms to estimate the channel based on these pilot symbols include minimum mean-square error (MMSE) and maximum-likelihood (ML) estimation. With a uniform density of pilots in frequency among the sub-carriers, the ML estimate exhibits a boundary effect whereby the quality of the estimates on sub-carriers towards the edges of the signal band degrades. While the use of MMSE mitigates this effect significantly it requires precise knowledge of channel statistics which are difficult to obtain. Also, it is often too complex to implement for common numbers of sub-carriers. This paper proposes mitigation of the boundary effect described above by transmitting extra pilot symbols in the sub-carriers towards the edges of the band in addition to the regularly spaced pilots, thereby increasing pilot density at the edges. Channel estimation is performed using a two-step algorithm. ML estimates are first formed using the uniform pilot set. The output of the ML estimator is then used to bias an MMSE estimator which is applied only to the edge pilot set. The two-step approach enhances performance as compared to ML estimation while keeping complexity much lower than that incurred when MMSE is used across the entire band.