Mixed-Timescale Channel Estimation for MIMO Relay Multi-User Systems Based on the PARAFAC Decomposition

This letter presents novel mixed-timescale channel estimation algorithms for uplink multiple-input multiple-output (MIMO) relay multi-user systems, considering the mixed-timescale property of practical dual-hop channel. With the one-stage training (OST) scheme, we formulate the channel estimation problem based on the parallel factor (PARAFAC) decomposition using the received pilots at multiple time slots, where the second-hop channel is viewed as a long-timescale variable and the first-hop channel is short-timescale. Then, the alternating least-square (LS) fitting method in batch mode (Batch-ALS) and online recursive alternating LS (Online-RALS) algorithms are proposed for mixed-timescale channel estimation. Simulation results show that the proposed algorithms improve the estimation performance significantly compared to the existing single-timescale channel estimation algorithms.