AR-model of faded MIMO set-up optimized for ITU channels

Autoregressive (AR) model based approximation for highly faded frequency selective time varying multi input multi output (MIMO) channels synthesized as per ITU specifications is investigated and reported in this paper. A semiblind approach is adopted for determining the channel state information (CSI) aided by training sequences and steepest gradient descent based on least mean square (LMS) adaptation. The AR model approximated for a given case is used to predict CSI of subsequent states. AR model order is derived for optimum performance of the system and a tradeoff is obtained between computational complexity and spectral efficiency of the system. Computational complexity for different cases of the channel is also discussed. Finally, signal to noise ratio (SNR) vs. bit error rate (BER) curve for different channel conditions for optimized AR model order is shown and certain conclusions drawn.