Performance evaluation of two-dimensional interpolations on OFDM channel estimation

Two-dimensional (2D) interpolation techniques for pilot-symbol-aided (PSA) OFDM channel estimation are investigated. The best linear interpolator/filter in terms of minimizing mean square error (MSE) is 2D Wiener filter. However, there are other (non-linear) 2D interpolation techniques such as nearest neighborhood (NN), bilinear, bicubic and 2D cubic spline which are frequently used in other applications such as in image processing. We apply these interpolation techniques for OFDM channel estimation in fast fading wireless channels and compare their performances with two-dimensional Wiener filtering. The suitable distance of pilot points are considered. We have also compared the performance of system for different Doppler frequencies.