Method to improve channel estimate in broadband simo/mimo cellular radio networks during abrupt interference variations

Abstract

A channel response matrix gathering all unknown elements for all K pilot subcarriers on links between transmitting and receiving antennas, is Least Squares (full-rank) estimated. The interference covariance matrix is estimated from the residual. The correlation of the interference covariance matrix is computed at each new symbol and compared with a threshold. If the interference is significantly changed, the interference covariance matrix is re-initialized, otherwise a running average of the interference covariance matrix is updated with the actual value. Both the updated/re-initialized interference covariance matrix and the autocorrelation matrix of the transmitted pilot sequences (stored in the receiver) are used for whitening the estimated channel matrix before submission to modal filtering for rank reduction. The original spatial and/or temporal correlations are reintroduced in the reduced-rank channel matrix by inverse whitening. After estimation in the discrete-time domain, DFT for conversion is applied for estimation in the whole frequency band.