Abstract— A computationally efficient optimal double correlation (ODC) filter, which works on the fourth order statistics of received samples, is proposed for carrier frequency offset estimation in MIMO OFDM systems using repetitive pilot symbols. The integral weights applied to the double correlation coefficients together with the requirement of a single phase operation allow particularly simple hardware implementation. Theoretical analysis shows that the mean square error in the estimate reaches the Cramer-Rao bound at high signal-to-noise ratios over a wide frequency range. Simulation results further indicate that the performance of the new algorithm is comparable to that of the maximum likelihood estimation solution even for large carrier frequency offsets.