Non-Coherent Multi-User Detection Based on Expectation Propagation

Ngo Khac-Hoang (1) , Maxime Guillaud (1) , Alexis Decurninge (1) , Yang Sheng (2) , Subrata Sarkar (3) , Philip Schniter (3)

1 Mathematical and Algorithmic Sciences Lab [Paris]
2 L2S - Laboratoire des signaux et systèmes
3 OSU - Ohio State University [Columbus]

In this paper, we propose a novel soft-output multi-user detector for non-coherent multiple access with Grassmannian signaling under Rayleigh block fading. Our detector is based on expectation propagation (EP) approximate inference and has polynomial complexity in the number of users. A simplified version of this scheme coincides with a scheme based on soft minimum-mean-square-error (MMSE) estimation and successive interference cancellation (SIC). Both schemes, especially EP, produce accurate approximates of the true posterior. They outperform a baseline decoder based on projecting the received signal onto the subspace orthogonal to the interference in terms of both hard-detected symbol error rate and coded bit error rate.

Domains: Engineering Sciences [physics]

Ngo Khac-Hoang, Maxime Guillaud, Alexis Decurninge, Yang Sheng, Subrata Sarkar, et al.. Non-Coherent Multi-User Detection Based on Expectation Propagation. 2019 53rd Asilomar Conference on Signals, Systems, and Computers, Nov 2019, Pacific Grove, United States. pp.2092-2096, ⟨10.1109/IEEECONF44664.2019.9049073⟩. ⟨hal-02556927⟩