2011/11/27 by Xuebing Han, Hao Zhang, Han, Xuebing +2
Computer Science · Engineering · #Blind Source Separation Techniques #FOS: Computer and information sciences #Image and Signal Denoising Methods #Information Theory (cs.IT) #Sparse and Compressive Sensing Techniques
paper · pdf · doi:10.48550/arxiv.1111.6237
openalex publication_date 2011/11/27 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
In this paper, we account for approaches of sparse recovery from large underdetermined linear models with perturbation present in both the measurements and the dictionary matrix. Existing methods have high computation and low efficiency. The total least-squares (TLS) criterion has well-documented merits in solving linear regression problems while FOCal Underdetermined System Solver (FOCUSS) has low-computation complexity in sparse recovery. Based on TLS and FOCUSS methods, the present paper develops more fast and robust algorithms, TLS-FOCUSS and SD-FOCUSS. TLS-FOCUSS algorithm is not only near-optimum but also fast in solving TLS optimization problems under sparsity constraints, and thus fit for large scale computation. In order to reduce the complexity of algorithm further, another suboptimal algorithm named D-FOCUSS is devised. SD-FOCUSS can be applied in MMV (multiple-measurement-vectors) TLS model, which fills the gap of solving linear regression problems under sparsity constraints. The convergence of TLS-FOCUSS algorithm and SD-FOCUSS algorithm is established with mathematical proof. The simulations illustrate the advantage of TLS-FOCUSS and SD-FOCUSS in accuracy and stability, compared with other algorithms.