Correlated Channel Feedback Method using Principal Component Analysis

A receiver transforms correlated channel state information (CSI) to an uncorrelated sparse CSI vector by using a principal component analysis (PCA), and feeds it back to a transmitter. The PCA extracts essential information from the CSI without redundancy that arises from the highly correlated antennas. The transmitter then recovers the original CSI through the inverse transformation of the feedback vector. To this end, we derive formally a covariance matrix of spatially correlated Rayleigh fading channels with its statistics including spatial correlation factors, channel variance, and channel delay profile. With only the knowledge of channel statistics, the transceiver can readily obtain the transformation matrix and employ a PCA based compression for CSI feedback; therefore, huge signaling overhead for the feedback can be reduced by about one quarter without compromising performance. Keywords—channel feedback; correlated channels; principal component analysis (PCA); compression