Reconstruction of low-rank jointly sparse signals from multiple measurement vectors
- 44 Downloads
The multiple measurement vectors problem approximates a set of signals sharing a common sparsity pattern simultaneously, using different linear combinations of those signals obtained through a sensing matrix. In a situation where the signal matrix has full row-rank, MUltiple SIgnal Classification (MUSIC) algorithm guarantees to recover the jointly sparse signals, but for the rank-defective case, the MUSIC performance is voided. To address such a rank deficient case, our proposed method, line search low-rank jointly sparse signals (LS-LRJSS), provides a geometric analysis of the problem by characterizing the linear dependence of the measurements with the linear coefficients that permit the reconstruction of each point from its neighbors. Moreover, a subspace analysis has been done on a Grassmann manifold to obtain the subspace that the signal matrix belongs to. Several numerical experiments evidence that the proposed method is more accurate and less time-consuming compared to existing approaches especially wherever the sparsity level of the signals increases or the number of measurement vectors decreases.
KeywordsCompressive sensing Multiple measurement vectors Joint sparsity Rank deficiency
- 19.Kreutz-Delgado, K., Rao, B.D.: Measures and algorithms for best basis selection, In: Proceedings of the 1998 IEEE International Conference on Acoustics, Speech and Signal Processing, vol. 3, pp. 1881–1884, IEEE (1998)Google Scholar
- 21.Makhzani, A., Valaee, S.: Reconstruction of jointly sparse signals using iterative hard thresholding. In: IEEE International Conference on Communications (ICC), pp. 3564–3568: IEEE (2012)Google Scholar
- 22.Foucart, S.: Recovering jointly sparse vectors via hard thresholding pursuit. In: Proceeding Sampling Theory and Applications (SampTA)] (2011)Google Scholar
- 26.Feng, P., Bresler, Y.: Spectrum-blind minimum-rate sampling and reconstruction of multiband signals. In: IEEE International Conference on Conference Proceedings on Acoustics, Speech, and Signal Processing, ICASSP-96, vol. 3, pp. 1688–1691: IEEE (1996)Google Scholar