Variational Inference-Based Iterative Receiver for Unified Non-Orthogonal Waveform (uNOW)

Non-orthogonal waveform has been identified as a promising spectrally efficient technology in the next generation wireless communications. In this correspondence, we develop computationally efficient iterative receivers from a unified variational inference perspective for unified non-orthogonal waveform (uNOW) signaling over multipath channels. Building on the constructed multi-layer factor graph, the parametric message passing algorithms for equalization are derived by invoking the mean field (MF) and Bethe approximation. To improve the convergence rate of the MF method, we propose its refined version with sequential message passing schedule to enhance the message updating. For Bethe approximation considering the dependence among symbols, we further propose its reduced-complexity version by exploiting Gaussian approximation. Simulation results demonstrate the benefits of the proposed variational inference-based message passing receivers conceived for uNOW signaling.