Persistently active block sparsity with application to direction-of-arrival estimation of moving sources

In this paper, the problem of recovering inconsistent sparse models from multiple observations is considered. A new method is developed by introducing a novel objective function, which exploits both block-level and element-level sparsities and promotes persistence in activity within a block. Then, we use a SVD-based method to reduce its computational complexity. Application of the method to the Direction-Of-Arrival (DOA) estimation of moving sources using a sensor array is presented and a simulation example is shown as a demonstration of the promising performance of the method in a moving DOA setting, particularly when sources are very close to each other.