A Least-squares Algorithm for Multipath Estimation Using an UWB-IR Link

This paper introduces a modified complex to real amplitude least-squares (CRALS) algorithm for the estimation of dense multipath in an UWB-IR channel. The traditional CRALS is ineffective for UWB-IR signal due to its wide bandwidth and large off period. This algorithm is modified to resolve three closely spaced paths in a correlation window. One novelty of the modified algorithm is that region-based processing is used for the initial estimate in place of coordinate descent processing used in traditional algorithm, and so the long iteration due to large off period is saved. The other novelty is that the procedures are simplified because of no necessity to increase the penalty term to infinity in UWB-IR case. Its efficiency is compared to the generalized maximum-likelihood estimator (GML) by simulation in different signal to noise ratios (SNR). The reason for its efficiency improvement comes from fewer effective spectrum samples and region-based processing of correlation results. Real measurement is conducted to verify this algorithm's performance. Its estimation error of time difference of arrival (TDOA) is about 0.18ns.