Fast-fading channel estimator using DCT and simplified kalman filter

We present a novel channel estimator for fading channels using basis function approximations and Kalman filtering. The proposal results in a low complexity estimator that is robust to the shape of the channel Doppler spectrum and allows different Doppler spectra for each channel coefficient in a frequency-selective scenario. The estimator is based on the description of channel variations via truncated discrete cosine transform (DCT). We show that DCT can be approximated by a set of linear narrow-band passband filters that can be inserted into a Kalman filter formulation that track a set of decoupled parameters. This estimator attains close to perfect channel knowledge performance with only 1% training symbols in a data frame with Doppler frequencies up to 0.8 times the frame frequency.