Blind source separation combining independent component analysis and beamforming
EURASIP Journal on Applied Signal Processing
IEEE Transactions on Audio, Speech, and Language Processing
A general modular framework for audio source separation
LVA/ICA'10 Proceedings of the 9th international conference on Latent variable analysis and signal separation
| Hi-index | 0.00 |
This work presents a novel robust method for a two-channel multiple Time Difference of Arrival (TDOA) estimation. The method is based on a recursive frequency-domain Independent Component Analysis (ICA) and on the novel State Coherence Transform (SCT). ICA is computed at different independent time-blocks and the obtained demixing matrices are used to generate observations of the propagation model of the intercepted sources. For the assumed time-frequency sparse dominance of the recorded sources, the observed propagation models are likely to represent all the active sources. The global coherence of the models is evaluated by a cumulated SCT, which provides a precise TDOA estimation for all the sources. Experimental results show that an accurate localization of 7 closely-spaced sources is possibile given only few seconds of data even in the case of low SNR. Experiments also show the advantage of the proposed strategy when compared with other popular two-microphone GCC-PHAT based methods.