we introduce a new method for sparse decomposition. Sparse decomposition aims to find sufficiently sparse solutions of a linear system of equations. As an application of SCA we consider our problem in context of blind source separation. In blind source separation problem, objective is to recover sources from their linear mixtures (observations where mixing matrix is also unknown. Our main interest is in underdetermined case, where mere knowledge of mixing matrix is not enough for source separation. We will define “active subspace” (linear subspace spanned by submatrix correspondding to active source components at a given time) and develop an iterative algorithm that detects active subspace and use the solution with minimal A2 norm in current active subspace to
redefine active subspace for next iteration. Experiments show that introduced method compromises between convergence speed and accuracy of well-known algorithms available in literature
