Abstract:

Given that the channel estimation algorithm which combines wavelet de-noising with interpolation in DFT cannot eliminate the noise in the cyclic prefix length, we propose a new channel estimation algorithm, a combination of wavelet de-noising and the improved DFT interpolation. This channel estimation algorithm first uses the discrete wavelet transform of least squares (LS) method to estimate the results of threshold de-noising processing, and sets the corresponding threshold in the process of DFT interpolation based on the noise variance of the mean inside and outside cyclic prefix. Then the noise of the cyclic prefix length is handled again to reduce further noise influence. Simulation results show that under the premise of unchanged complexity, the proposed algorithm cannot only remove the influence of additive white Gauss noise better, but also improve the accuracy of channel estimation effectively. In general, the channel estimation algorithm performance is better than the combination of wavelet de-noising and DFT interpolation.

Key words: OFDM, channel estimation, wavelet de-noising, DFT