Paper No: 128
Two improved sparse decomposition methods for Blind Source Separation
Author(s): B. Vikrham Gowreesunker, Ahmed H. Tewfik
In underdetermined Blind Source Separation problems, it is common
practice to exploit the underlying sparsity of the sources for
demixing. In this work, we show how trained dictionary can improve
sparsity of the underlying sources. We also propose two sparse
decomposition algorithms for the separation of linear instantaneous
speech mixtures. The first algorithm uses a single channel Bounded
Error Subset Selection (BESS) method to estimate the mixing matrix.
The second sparse decomposition method performs a constraint
decomposition of the mixtures over a stereo dictionary.