Local Linear Regression Based Channel Estimation for OFDM Systems

A local linear regression (LLR) based least-square (LS) channel estimation technique for mobile orthogonal frequency division multiplexing (OFDM) communication systems over the doubly selective fading channel is proposed in this paper. In our proposed method, the doubly selective fading channel is firstly estimated via pilots arranged in block-type, comb-type or distributed pilot scheme, piece-wise linear interpolated and at last smoothed by local linear regression. The last smoothing procedure greatly decrease the estimation error introduced by channel noise and interpolation error so achieved better mean square error (MSE) performance and symbol error rate (SER) performance than the conventional interpolation method such as linear interpolation and DFT based interpolation, and the error floor is greatly decreased compared to DFT based interpolation scheme. The computational complexity increased by the smoothing procedure is only linearly proportional to the effective subcarrier-number. Simulations show the performance improvements offered by our approach to the existing ones.