Improved Algorithm Based on the DFT Transform Channel Estimation and Threshold Filtering

Abstract

In channel estimation, the complexity of the DFT-based channel estimation algorithm is lower than the MMSE algorithm, and the DFT-based algorithm performs better than the LS algorithm. However, due to that the traditional algorithms simply considered all the samples to be useful channel impulse response, ignoried the effect of noise. Therefore, the algorithm could be improved. This paper presents an improved algorithm based on twice noise estimation theory. The algorithm uses the sequence of points other than the cyclic prefix length to estimate the noise variance firstly, and then use the estimated noise variance to distinguish the noise samples within the cyclic prefix length. The computation of the new noise variance and noise mean through the new noise points as a threshold filters the impulse response of the channel in time domain and eliminates the impact of noise on the system further. The simulation result shows that the improved algorithm performs better than the traditional algorithm.