Scattered Pilot Design And Channel Estimation In OQAM/FBMC Communication Systems

Filter bank multicarrier with offset quadrature amplitude modulation(OQAM/FBMC) is a promising alternative to the conventional orthogonal frequency division multiplexing(OFDM) technique. In the recent past, OQAM/FBMC and other alternative filter bank-based waveforms that can overcome the shortcomings of OFDM in multiuser networks have been studied.However, channel estimation in OQAM/FBMC, in general, is more complicated than in OFDM. Scattered pilots are commonly inserted among data symbols in the time-frequency space of an OFDM waveform for channel estimation and its tracking over the length of each data packet. The orthogonality of subcarriers in OFDM allows separation of pilot subcarriers from data subcarriers and, hence, channel estimation at the pilot positions can be performed trivially. Unfortunately, this is not the case in OQAM/FBMC, as subcarriers orthogonality is limited to the real field only(not the complex field).In this paper, we first propose a novel scattered pilot method for channel estimation in OQAM/FBMC systems. We refer to this as coded auxiliary pilot(CAP) method. It builds on the idea of using a few auxiliary pilot(AP) symbols to eliminate the imaginary interference on each scattered pilot. At the same time, to keep the spectral efficiency of transmission high, the AP symbols are coded to carry information symbols as well. Theoretical analysis and numerical results show that the CAP method can maintain high spectral efficiency as well as high power efficiency, while achieving complete imaginary interference cancellation.Based on the CAP shceme, in this paper, we also develop an iterative channel estimation and refinement algorithm to deal with the highly frequency selective channel, building on a recently proposed frequency spreading FBMC receiver structure. Simulation results reveal that the proposed algorithm obtains channel estimates whose estimation error follows the Cramer-Rao bound(CRB) very closely for signal-to-noise(SNR) values of as high as 30 dB.