Skip to Main Content
In this letter, the problem of blind symbol-timing and frequency-offset estimation in orthogonal frequency-division multiplexing (OFDM) systems with real data symbols is considered. Maximum-likelihood (ML) estimators for the parameters of interest have been derived by modeling the OFDM signal as a proper complex Gaussian random process. However, when data symbols belong to a real constellation, the received signal becomes an improper complex random process, and hence, previously mentioned estimators, termed cyclic prefix (CP) estimators, are not ML estimators. In this letter, minimum mean-squared error estimators exploiting the conjugate-symmetry property exhibited by the OFDM signal with real data symbols, termed MSR estimators, that greatly outperform CP estimators in additive noise channels, are derived. Moreover, a modified MSR symbol-timing estimator that can assure satisfactory performance in multipath fading channels is proposed.
Date of Publication: Oct. 2004