Rao-Blackwellized Gauss-Hermite filter for joint frequency offset and channel estimation for the MIMO-OFDM system

We propose a new frequency offset and channel estimation algorithm for a MIMO-OFDM system. To estimate offset parameters which enter the observation function in a nonlinear manner, we propose a Gauss-Hermite Filter which incorporates the Kalman filter (KF) conditioned on a random offset sample. To reduce the complexity, the Rao-Blackwellization is employed. In the MIMO-OFDM system, the initial estimate for the offsets and channels are obtained using one training symbol. In the remaining OFDM symbol intervals, the MIMO data detector is employed to detect the transmitted data symbols. The used data detector and Gauss-Hermite filter based channel/offset estimator are finally combined in a joint decision-directed algorithm.