Improved channel estimation using noise reduction for MB-OFDM UWB systems

The channel estimation is a crucial issue for ultra-wideband (UWB) systems based on the scheme of multi-band OFDM (MB-OFDM), where coherent detection is employed in a general way. The accuracy of channel identification is limited by noise in a certain extend, therefore, measures for noise reduction should be taken. This paper investigates how the performance of a least square (LS) channel estimator can be improved by estimating and suppressing the noise in the signal subspace, according to the property of the UWB channel in a MB-OFDM system. This algorithm of noise reduction is based on the processing of correlation between the noise within and outside the channel impulse response (CIR) subspace. Simulation results show that better performance can be achieved for a range of signal-to-noise ratio (SNR).