CS-based device-free localization in the presence of model errors

Abstract

Compressive sensing (CS) has recently been applied for device-free localization (DFL) by exploiting spatial sparsity to reduce the number of measurements required by DFL systems while maintaining the high localization accuracy. However, few works considered model errors in CS-based DFL. This paper proposes an adaptive sparsity-based DFL appoach to overcome the problem incurred by model errors. The novel feature of this method is to dynamically adjust the basis matrix (a.k.a. dictionary) based on a two-stage dictionary learning (DL) framework with non-negativity constraints. Compared to previous CS-based DFL methods, the proposed method can compensate the inaccuracy of the basis matrix and improve sparse reconstruction performance simultaneously. Experimental results verify the performance of the proposed approach on the location accuracy.