A PCA-based modeling method for wireless MIMO channel

Geometry-based stochastic model (GBSM) of multiple input multiple output (MIMO) channel describes the channel impulse response (CIR) in the sense of rays and clusters, which obey the empirical distributions. Thus, the correlation between MIMO sub-channels is not explicitly defined, which makes it difficult for GBSM to predict channel capacity accurately. Facing the increased antenna number of massive MIMO for fifth generation (5G) communication, we propose a channel modeling method using principal component analysis (PCA). This method takes advantage of the hidden features and structures extracted from the measured channel data, combining the information of the scenario and antenna configurations, to reconstruct the amplitude and phase of the CIR respectively. The sparse features and structures can efficiently support the high antenna number of massive MIMO. By the proposed scheme, the accuracy of 56 × 32 MIMO capacity improves 12.8% compared with the GBSM model.