On blind CFO estimation for FBMC-PAM systems

This paper deals with the problem of blind carrier-frequency offset (CFO) estimation for filter bank multicarrier systems with pulse amplitude modulation (FBMC-PAM). In FBMC-PAM a sequence of real-valued symbols can be transmitted over 2 M spectrally separated subcarriers operating at the same symbol rate as OFDM. This recently proposed system does not require a cyclic prefix and, in the multi-user context, it provides a high level of spectral separation among users. In this paper is derived the maximum likelihood (ML) blind CFO estimator for AWGN channel under the assumption of low SNR conditions. Since the FBMC-PAM signal is a noncircular random process, the obtained ML estimator exploits both the unconjugate and the conjugate correlation (or relation function). The performance of the derived CFO estimator is compared with that of the two estimators obtained by separately maximizing the contribution to the ML cost function exploiting the unconjugate correlation or the conjugate correlation, and with the modified Cramér-Rao bound.