UKF-based channel estimation and LOS/NLOS classification in UWB wireless networks

Abstract

The paper addresses ultra-wideband (UWB) channel estimation and line-of-sight (LOS) vs. non-line-of-sight classification based on the application of the unscented Kalman filter (UKF) and the analysis of the multipath channel response characteristics. For non-linear models, the UKF provides an efficient recursive minimum mean squared error estimation technique that is successfully applied in this work, and supported by numerical results demonstrating its effectiveness in UWB multipath channel gain and time delay estimation. The multipath channel response obtained from a limited number of channel taps is subsequently characterized in terms of relevant statistical parameters including kurtosis and mean excess delay. This characterization reveals clear differences in the statistics of these parameters under LOS and NLOS propagation conditions for various channel types in residential, office, outdoor and industrial environments. Based on the estimated parameters probability density functions under LOS/NLOS conditions, a likelihood ratio test (LRT) for hypothesis classification is performed for the different UWB channel models, and numerical results show that highly reliable LOS vs. NLOS classification is achievable, with accuracy exceeding 90% for most cases of practical interest. These results can then be further exploited in enhancing the performance of positioning applications.