An Efficient Algorithm to Estimate Mixture Matrix in Blind Source Separation using Tensor Decomposition

The estimation of mixing matrix is a key step to solve the problem of blind source separation. The existing algorithm can only estimate the matrix of well-determined, over-determined and under-determined in condition of sparse source. Scaling and permutation ambiguities lie in both factor matrix of tensor Canonical Decomposition and mixing matrix in blind source separation. With this property, the estimation of mixing matrix can be transformed into tensor Canonical Decomposition of observed signal’s statistic. The decomposition can be realized by the method of alternating least squares. The theoretical analysis and simulations show that the method proposed in this paper is an efficient algorithm to estimate well-determined, over-determined and under-determined mixing matrix.