A Noise Variance Optimization Method for 2x1-Dimensional Wiener Filtered Channel Estimation

A noise variance optimization method is proposed for the time and frequency dimension separate (2 times 1D) Wiener-filtered channel estimation of OFDM based systems. According to Wiener-filter theory, the noise variance is necessary to achieve optimal solution. For 2 times 1D Wiener-filtered channel estimation, the Wiener-filtering will be applied twice respectively in time and frequency dimension. However, the effect of variety of noise variance induced by the first filter should be considered on the second filter in this method. In the proposed method, the noise variance used by the second filter is optimized according to the mean square error (MSE) of channel estimation by the first filter. The exact MSE of channel estimation is derived in this paper. Moreover, the channel estimation performance is evaluated with different noise variance optimizing criteria. The simulation results show that the performance of the proposed method is better than the 2 times 1D filters method without noise variance optimization, and is very close to that of the Wiener 2D filter.