LAUSR

For proper matrix ensembles, it has been known that the greedy pursuit (GP) algorithms are computationally efficient and fast to reconstruct sparse signals from far fewer linear measurements. In considering several parameters such as sparsity level, sparse signal ambient dimension and the number of linear measurements, the GP algorithms have been shown to perform differently in estimating sparse signals. According to data fusion principle, fusing completely the estimated support set of different reconstruction algorithms can improve signal recovery performance. It can, however, lead to the increased probability of estimating incorrect support indices, and thus degrades the signal reconstruction accuracy. In this paper, a new fusion framework, namely collaborative framework of algorithms (CoFA), is proposed to pursue accurate reconstruction of the sparse signals from far fewer linear measurements. The two main ingredients of the proposed scheme that control the estimation of incorrect support indices are pre-selection support of orthogonal matching pursuit (OMP) algorithm and Thresholding -to eliminate unpromising indices from the identied support set of any participating algorithm. Using the restricted isometry property, the theoretical analysis of the CoFA scheme and the sufficient conditions (guarantees) for realizing an improved reconstruction performance are presented. Simulation results demonstrate that the proposed scheme is effective and offer a better channel estimation performance in terms of mean-squared-error (MSE) and bit-error-rate (BER) when compared to other reconstruction algorithms, without the significant increase in computational complexity.