Abstract
AbstractWe address the problem of blind source separation in the underdetermined mixture case. Two statistical tests are proposed to reduce the number of empirical parameters involved in standard sparseness-based underdetermined blind source separation (UBSS) methods. The first test performs multisource selection of the suitable time–frequency points for source recovery and is full automatic. The second one is dedicated to autosource selection for mixing matrix estimation and requires fixing two parameters only, regardless of the instrumented SNRs. We experimentally show that the use of these tests incurs no performance loss and even improves the performance of standard weak-sparseness UBSS approaches.
Highlights
Source separation is aimed at reconstructing multiple sources from multiple observations captured by an array of sensors
We show that our statistical algorithms reduce the number of empirical parameters and improve the overall performance of the underdetermined blind source separation (UBSS) methods under consideration
The modified signal norm testing (SNT)-time–frequency ratio of mixtures (TIFROM) performs multisource selection and forces to zero the unselected time–frequency points. These results show that SNT makes it possible to select the autosource time–frequency points, with no performance loss and without resorting to the empirical threshold required by the original TIFROM
Summary
Source separation is aimed at reconstructing multiple sources from multiple observations (mixtures) captured by an array of sensors. M j=1 σj Signal source detection for mixing matrix estimation (autosource selection) we propose a test for selecting the time– frequency points where one signal source is probably present alone To perform this selection, we make the distinction between signals with either low or high overlapping rate in the time–frequency domain. SUBSS method The modified SUBSS algorithm is obtained by using both the DATE and SNT for source recovery and mixing matrix estimation by SNT, respectively, as explained in Section “Statistical tests for sparseness-based UBSS”. The gain brought by the multisource selection, which acts as a denoising, is bigger on a wider SNR range because the time–frequency representation of chirp signals is sparser than that of audio signals
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have