Beam Domain Channel Estimation in Spatial Non-Stationary Massive MIMO

The patent applies local quality by allowing different scattering clusters to have different visibility regions and sparse structures. Specifically, wholly visible (WV) clusters have full array visibility while partially visible (PV) clusters have limited visibility, creating location-dependent channel characteristics. This local differentiation resolves the contradiction by enabling low pilot overhead through sparsity exploitation while maintaining accurate estimation through non-stationary modeling.

The patent changes the parameter of channel sparsity structure from a common fixed structure to a variable structure that adapts to different scattering clusters. By introducing visibility region parameters and cluster-specific sparse patterns, the method dynamically adjusts the sparsity model to match actual channel conditions, achieving both low pilot overhead and high estimation accuracy.

The patent segments the channel estimation process into distinct stages: first identifying dominant beams through coarse search, then refining beam support sets by considering power leakage and spatial non-stationarity. This segmentation allows the algorithm to achieve high accuracy without excessive complexity by processing different aspects of the channel model in separate, manageable steps.

The patent introduces dynamic adaptation in the beam domain estimation by adjusting the power ratio threshold based on observed power leakage patterns. The algorithm dynamically refines the beam support set rather than using a fixed threshold, enabling accurate estimation while maintaining reasonable computational complexity through adaptive rather than exhaustive processing.