 |
In this talk we consider a class of sparse Bayesian algorithms for the estimation of wireless channels in the context of their application to localization. In a nutshell, these algorithms aim at detecting and estimating dominant “specular-like multipath components” in the channel response. Specifically, the number of said components and their parameters, such as their relative delay and complex amplitude, are estimated. This information can be exploited for localization purpose, e.g. by finger-printing or by reconstructing the corresponding physical propagation paths between transmitter and receiver, as done in SLAM.
We discuss the key properties of these algorithms such as their ability to detect components and to resolve them in the dispersion domain (e.g. with respect to their relative delay). We also shed some light on the correct interpretation of “components” extracted by such algorithms (and actually by any parametric algorithm). We discuss the implications of these properties on localization schemes based on multipath reconstruction.
- Poster
