An Improved DFT-Based Channel Estimation Algorithm for OFDM System in Non-sample-Spaced Multipath Channels

It is well known that channel impairment caused by multipath reflections can deeply degrade the transmission efficiency in wireless communication systems. The conventional DFT-based channel estimation methods improve the performance by neglecting nonsignificant channel taps. However, in multipath channels with non-sample-spaced time delays, this will cause power leakage and result in an error floor. In order to overcome this problem, based on the property of Channel impulse response, we describe a utility of channel estimation technology for OFDM systems using discriminant analysis. Usually, significant channel taps are detected on the basis of a predetermined threshold, so the optimal threshold value becomes a crucial factor. But it is difficult to decide channel taps that are approximately equal to the threshold value. To solve this disadvantage, the proposed algorithm improves performance by using Mahalanobis distance discriminant analysis for channel taps. This approach can mitigate the aliasing effect in the DFT-based channel estimator and reduce the leakage power efficiently when there is non-sample-spaced path delay. Simulation results show that the proposed channel estimators can eliminate the error floor substantially and its performance is better than the LS and conventional DFT estimators.