Channel Tracking using Pruning for MIMO-OFDM Systems over Gauss-Markov Channels

In this paper we investigate the problem of channel tracking and detection for MIMO-OFDM systems over fast varying channels obeying a Gauss-Markov model. We consider time domain tracking of the channel matrix taps with Kalman filter, whereas symbols detection is carried out by a zero-forcing (ZF) soft detector. A key assumption of the theory of Kalman filter is that the state-space model is perfectly known, while communication systems make use of the detected symbols as an input to the Kaiman filter in order to form a suitable state-space model. This gives rise to error propagation due to misdetected symbols (model mismatch) and is usually solved by using frequently inserted pilot symbols, resulting in a reduced spectral efficiency. To overcome this problem, we suggest a novel approach to mitigate the error propagation due to misdetections without using frequent pilot symbols. In particular, we consider the reliability of the detections based on the soft detector and use only those outputs that have robust reliability to track the channel matrix taps, minimizing the effect of Kalman filter mismodeling. This method can significantly reduce the error propagation effect, leading to an improved bit error probability.