Optimal MMSE finite parameter model for doubly-selective channels

Multipath propagation and mobility-induced Doppler effects contribute to the time- and frequency-selectivity (doubly-selectivity) in channels. To describe the doubly-selective channel, we propose a finite parameter model composed of kernels expressed as a function of the channel correlation based on Karhunen-Loeve (KL) expansion. The proposed model is optimal in the MMSE sense, provided that the instantaneous channel correlation is matched. It can also adapt to changes in the velocity of the terminal. Even when the channel correlation is unavailable and can only be assumed from Jakes´ model, the channel estimation mean square error (MSE) of our proposed model is shown to be comparable or outperforms the oversampled basis expansion model (OBEM), polynomial model (PM) and Slepian model as validated by Monte-Carlo simulations on Rayleigh fading channels. In the application in an OFDM system, the proposed model confers significant BER performance gain over OBEM and PM