Compressive sensing based target counting and localization exploiting joint sparsity

One of the fundamental issues in Wireless Sensor Networks (WSN) is to count and localize multiple targets accurately. In this context, there has been an increasing interest in the literature in using Compressive Sensing (CS) based techniques by exploiting the sparse nature of spatially distributed targets within the monitored area. However, most existing works aim to count and localize the sparse targets utilizing a Single Measurement Vector (SMV) model. In this paper, we consider the problem of counting and localizing multiple targets exploiting the joint sparsity feature of a Multiple Measurement Vector (MMV) model. Furthermore, the conventional MMV formulation in which the same measurement matrix is used for all sensors is not valid any more in practical time-varying wireless environments. To overcome this issue, we reformulate the MMV problem into a conventional SMV in which MMVs are vectorized. Subsequently, we propose a novel reconstruction algorithm which does not need the prior knowledge of the sparsity level unlike the most existing CS-based approaches. Finally, we evaluate the performance of the proposed algorithm and demonstrate the superiority of the proposed MMV approach over its SMV counterpart in terms of target counting and localization accuracies.