Cluster-Based Multi-Target Localization Using Joint Sparsity Property

In the multi-target localization scenario, adopting the compressive sensing technique can effectively reduce the acquisition costs by collecting only a small number of measurements to recover the target-position signal. This signal reveals the locations of the targets and is evidently sparse due to the spatial sparseness of point targets. However, the measurement noise may seriously degrade the localization performance. To improve the localization accuracy, this paper proposes a hierarchical multi-target localization approach using a cluster-based sensor network. Therein, several cluster heads collect and preprocess the measurements from their neighboring sensor nodes and then report these intra-cluster measurements to a fusion center which finally recovers the position signals. Solutions representing the measurements collected at all cluster heads are modeled to possess a useful mathematical property, namely the joint sparsity property. Accordingly, a row-based least absolute shrinkage and selection operator formulation is established to utilize this joint sparsity property. By doing so, the cooperation gains among all clusters are efficiently collected against the noise effect, through joint signal reconstruction. Simulation results show that the proposed cluster-based multi-target localization approach improves the localization accuracy over non-cluster approaches. Further, the simulation evaluation sheds light on the design guideline for the proposed approach implementation. That is, it can work well with a moderate number of clusters and a reasonable size of clusters.