Deep Learning-Based Channel Estimation and Equalization Scheme for FBMC/OQAM Systems

Filter bank multicarrier (FBMC) modulation is a promising candidate modulation method for future communication systems. However, FBMC systems cannot directly use channel estimation methods proposed for orthogonal frequency-division multiplexing systems due to its inherent imaginary interference. In this letter, we propose a channel estimation and equalization scheme based on deep learning (DL-CE) for FBMC systems. In the DL-CE scheme, the channel state information and the constellation demapping method are learned by a deep neural networks model, and then the distorted frequency-domain sequences are equalized implicitly to obtain binary bits directly. Simulation results show that the proposed DL-CE scheme achieves state-of-the-art performance on channel estimation and equalization.