Extended target localization using the variational Garrote

摘要

In several high-resolution array processing applications such as radar and sonar, it is necessary to localize targets that have a finite angular spread. In such scenarios, conventional subspace-based techniques tend to provide erroneous results, and hence, an extended target model is more appropriate. In this work, we consider a multiple input multiple output system model and propose to jointly estimate the range, Doppler shift and angular spread of extended targets. More specifically, we show that the extended nature of the target leads to a block-sparse recovery problem. To solve the problem, we present an extension of the variational Garrote method to estimate the unknown block sparse vector. The variational Garrote provides a simple approach for direct feature subset selection via a variational approximation to the posterior distribution over the subsets. In addition, our proposed method also takes into account the scaling of the angular spread of the target(s) at different distances. We illustrate the efficacy of our approach using Monte Carlo simulations.