Decision-Directed MIMO Channel Tracking with Efficient Error Propagation Mitigation

We treat decision-directed tracking of fast-varying MIMO channels by applying the Kalman filter. The Kalman filter is considered but the suggested algorithm is applicable independently of the tracking technique. In the decision-directed mode, tracking filters suffer from error propagation caused by wrongly detected data. To mitigate the error propagation, a recently suggested pilot on request training is applied. Appropriate metrics to detect the filter divergence were developed in the past but these metrics strongly depend on the tracking algorithm. In this paper, we design a metric which solely depends on the estimated channel. The metric threshold is derived analytically and is shown to be SNR-independent on a large SNR range. We show the effectiveness of the pilot on request training with the novel metric in both reducing the BER and saving bandwidth.