Maximum-Likelihood Timing/Channel Estimation for OFDM Systems with Cyclic Training Symbols

We present a new algorithm for joint timing and channel estimation in OFDM systems with cyclic training symbols. The proposed algorithm employs multiple signal classification (MUSIC) approach for coarse timing estimation and then refines the timing/channel estimation by applying the generalized maximum likelihood rule (GMLR) within a symbol period around the coarse timing for different channel lengths. We reveal that the likelihood corresponding to different shift positions and channel lengths can be conveniently computed from a single N-length channel estimate, which considerably reduces the computational complexity and thus makes the refined search over long training symbols feasible.