Subspace system identification for training-based MIMO channel estimation

The application of state-space-based subspace system identification methods to training-based estimation for quasi-static multi-input-multi-output (MIMO) frequency-selective channels is explored with the motivation for better model approximation performance. A modification of the traditional subspace methods is derived to suit the non-contiguous nature of training data in mobile communication systems. To track the time variation of the channel, a new recursive subspace-based channel estimation is proposed and demonstrated in simulation with practical MIMO channel models. The comparison between the state-space-based channel estimation algorithm and the FIR-based Recursive Least Squares algorithm shows the former is a more robust modeling approach than the latter.