Diagonally Loaded Linear MMSE Equalization for UWB Multipath Channel Under Limited Training Sample Support

In this paper, diagonally loaded linear minimum mean square error (MMSE) equalization technique is proposed for ultra wideband (UWB) multipath channel in the presence of inter symbol interference (ISI) and multiple access interference (MAI). In the case of practical implementation under limited training sample support, the optimal solution for linear least squares problems is the minimum norm (MN) solution. But it suffers from performance degradation when the sample covariance matrices are ill-conditioned. Diagonal loading (DL) technique is incorporated into MN solution and shown to achieve significant performance improvement when the amount of training symbols is comparable with the number of taps utilized by the equalizer. Moreover, a rank-reduced multistage Wiener filter (MWF) with DL is proposed for UWB multipath channel equalization which requires large number of taps. It is shown to converge to the MN solution with much reduced computational complexity