We are revisiting the problem of Blind Source Separation and Blind Source Extraction (BSS/BSE) based on the assumption that original signals in mixtures are independent. In BSE, the goal is to extract one or more sources of interest (SOIs). In BSS, the goal is to separate all signals in the mixture.
This page collects our recent outputs, that is, algorithms, codes of simulations, examples, etc.
Software outputs of Z. Koldovský, V. Kautský, P. Tichavský, J. Čmejla and J. Málek, “Dynamic Independent Component/Vector Analysis: Time-Variant Linear Mixtures Separable by Time-Invariant Beamformers, IEEE Trans. on Signal Processing, vol. 69, pp. 2158-2173, March 2021. (here)
FastDIVA package (Matlab implementation of the algorithm, demo example); FastDIVA in an extension of the famous FastICA for BSS/BSE of moving sources; it is mainly based on the mixing model with constant separating vector (CSV); one-unit, symmetric and block-deflation variants are implemented; it is applicable to real-valued as well as complex-valued mixtures; joint BSS/BSE of mixtures similar to IVA/IVE is possible as well; when data are considered as one block only, FastDIVA corresponds to FastICA in the real-valued case and is similar (simpler and not constrained to circular signal-of-interest) than the complex-valued FastICA/FastIVA in the complex-valued case:
 Download FastDIVA here |
 Report bugs here |
Matlab codes of examples from N. Amor, J. Čmejla, V. Kautský, Z. Koldovský, and T. Kounovský, “Blind Extraction of Moving Sources via Independent Component and Vector Analysis: Examples, ” to be presented at ICASSP 2021: here
Moving speaker extraction with dense microphone array: here (Please note that the results do not perfectly match with those in Table 1 in the corresponding paper due to some minor bugs we fixed after the paper submission. Nevertheless, the differences in the results are negligible and do not change our conclusions about the experiment.)
Moving activity in visual evoked potentials: here
Simulated block-by-block online extraction: here
Software outputs of Z. Koldovský, V. Kautský, T. Kounovský and J. Čmejla, “Algorithm for Independent Vector Extraction Based on Semi-Time-Variant Mixing Model, ” arXiv:1910.10242[eess.SP]:
Algorithm for dynamic blind extraction based on CSV suitable for on-line deployments: here
Software outputs of Z. Koldovský and P. Tichavský, “Gradient Algorithms for Complex Non-Gaussian Independent Component/Vector Extraction, Question of Convergence,”IEEE Transactions on Signal Processing, vol. 67, no. 4, pp. 1050-1064, Feb. 2019. arXiv:1803.10108[eess.SP]:
Gradient ICE Algorithms: here
Gradient IVE Algorithms (including a piloted version): here
Simulation of BSE comparing various ICE, IVE, BSE, ICA algorithms: here