Point Cloud Geometry and Attribute Transmission over MIMO Channels

Conventional point cloud delivery schemes employ tree-based or graph-based digital compression techniques to stream three-dimensional (3D) points and their associated attributes over wireless multiple-input multiple-output (MIMO) channels for 3D scene reconstruction. However, these digital-based delivery methods suffer from the cliff and leveling effects. Moreover, in MIMO communication environments, poor quality in some spatial layers, i.e., subchannels, can lead to point cloud quality degradation. We propose a novel point cloud delivery scheme that addresses both effects simultaneously. Our approach leverages a k-dimensional tree (k-d tree)-based graph Fourier transform (GFT) for energy compaction, a subchannel assignment mechanism to fully utilize subchannel diversity in MIMO links, and analog modulation with non-uniform power allocation for error-resilient and fluctuation-resilient transmission. Evaluation results demonstrate that the proposed scheme delivers superior 3D reconstruction quality compared to conventional digital-based point cloud delivery schemes, such as geometry-based point cloud compression (G-PCC), in noisy MIMO channels.