An active divide-and-conquer algorithm for sparse recovery support: Fluctuating targets case

We introduce a novel sparse signal detection algorithm, whose goal is to localize an unknown number of fluctuating targets in a sparse scenario, for example composed by N target sites. Rather than probing each cell with standard detection techniques, we employ an adaptive, step-by-step, tree-structured algorithm, that iteratively narrows the set of possible candidate locations. More specifically, assume that the transmitter is able to illuminate partitions of the target space at time and that the receiver acquires noisy nonlinear measurements of the partitions. Based on a comparison between these measurements and a threshold, the algorithm decides which partitions may contain targets (and, thus, they need to be probed at the next step) and which should be immediately discarded (as targets may not be present). Results show that the described algorithm is able to determine the location of the targets with fairly low False Discovery Proportion (FDP) and decreasing Non Discovery Proportion (NDP), as the available budget energy allocated to the algorithm increases.