OMP Based Joint Sparsity Pattern Recovery Under Communication Constraints (1307.8320v2)

Abstract: We address the problem of joint sparsity pattern recovery based on low dimensional multiple measurement vectors (MMVs) in resource constrained distributed networks. We assume that distributed nodes observe sparse signals which share the same sparsity pattern and each node obtains measurements via a low dimensional linear operator. When the measurements are collected at distributed nodes in a communication network, it is often required that joint sparse recovery be performed under inherent resource constraints such as communication bandwidth and transmit/processing power. We present two approaches to take the communication constraints into account while performing common sparsity pattern recovery. First, we explore the use of a shared multiple access channel (MAC) in forwarding observations residing at each node to a fusion center. With MAC, while the bandwidth requirement does not depend on the number of nodes, the fusion center has access to only a linear combination of the observations. We discuss the conditions under which the common sparsity pattern can be estimated reliably. Second, we develop two collaborative algorithms based on Orthogonal Matching Pursuit (OMP), to jointly estimate the common sparsity pattern in a decentralized manner with a low communication overhead. In the proposed algorithms, each node exploits collaboration among neighboring nodes by sharing a small amount of information for fusion at different stages in estimating the indices of the true support in a greedy manner. Efficiency and effectiveness of the proposed algorithms are demonstrated via simulations along with a comparison with the most related existing algorithms considering the trade-off between the performance gain and the communication overhead.