OFDM Channel Estimation by Adaptive Filtering

Abstract

The potentials of pilot symbol aided channel estimation are examined for OFDM based mobile radio and broadcasting applications. Channel estimation using two dimensional (2-D) Wiener filter could be replaced by two cascaded 1-D Wiener filters. The choice of two cascaded 1-D Wiener filters gives a good trade-off between performance and complexity. The two cascaded orthogonal 1-D filters are simpler to implement and the performance is almost the same as one 2-D Wiener filter. The optimum Wiener filter for pilot symbol aided channel estimation requires the perfect knowledge of the channel correlation function, which is unknown in practice. Conventional robust Wiener filter uses the model mismatch,i.e. it assumes a different correlation function with respect to the actual channel conditions. The conventional robust Wiener filter for estimation in time direction is designed for the maximum assumed Doppler-shift, and the filter in frequency-direction for the worst case channel delay spread. Often, the transmission channel behaves much better than the assumed worst case. In those situations, Wiener filters with a smaller bandwidth would yield better estimation results by reducing the noise on the channel estimate, but due to the model mismatch the robust Wiener filter can perform significantly worse for certain channel models than the optimum Wiener filter. In this thesis, we propose an improved channel estimator for a pilot based, mobile OFDM-system by adaptive Wiener-filtering to reduce performance loss with a moderate increase in complexity. The conventional fixed Wiener-filter is replaced by an adaptive filter, which adapts to the actual channel characteristic. The theory developed is then tested by simulations, and the analytical results are presented.