Blind signals separation with genetic algorithm and particle swarm optimization based on mutual information

Radioelectronics and Communications Systems (Impact Factor: 0.19). 06/2011; 54(6):315-324. DOI: 10.3103/S0735272711060045

ABSTRACT Blind source separation technique separates mixed signals blindly without any information on the mixing system. In this paper,
we have used two evolutionary algorithms, genetic algorithm and particle swarm optimization for blind source separation. In
these techniques a novel fitness function that is based on the mutual information and high order statistics is proposed. In
order to evaluate and compare the performance of these methods, we have focused on separation of noisy and noiseless sources.
Simulations results demonstrate that the proposed method for employing fitness function has rapid convergence, simplicity
and a more favorable signal to noise ratio for separation tasks based on particle swarm optimization and continuous genetic
algorithm than binary genetic algorithm. Also, particle swarm optimization enjoys shorter computation time than the other
two algorithms for solving these optimization problems for multiple sources.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: An enhanced blind source separation algorithm based on Stone’s BSS approach is proposed, to reject the Electrooculogram (EOG) artifact and power line noise (50Hz) from simulated and real human Electroencephalography (EEG) signals without the notch filter, in order not to lose any useful EEG data around the 50-Hz. The proposed algorithm which called efficient Stone’s BSS (ESBSS) has been compared with four well-known BSS algorithms over super- Gaussian, sub-Gaussian artifacts and EEG signals with a linear mixture. In Original Stone’s BSS, the half-life values taken as a constant, typically (h long ≥100 h Short), but in the proposed work, an optimization procedure is used to change these values until the maximum temporal predictability is found. The real EEG data are taken from Imperial College London using a computerized EEG device with eight electrodes placed according to the 10–20 system.
    Applied Mechanics and Materials 03/2014; 543-547:2687-2691. DOI:10.4028/
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: An efficient Stone’s BSS (ESBSS) algorithm is proposed based on the joint between original Stone’s BSS (SBSS) and genetic algorithm (GA). Original Stone’s BSS has several advantages compared with independent component analysis (ICA) techniques, where the BSS problem in Stone’s BSS is simplified to generalized eigenvalue decomposition (GEVD), but it’s susceptible to the local minima problem. Therefore, GA is used to overcome this problem and to enhance the separation process. Performance of the proposed algorithm is first tested through a different pdf source, followed by artifact extraction test for EEG mixtures then compared with the original Stone’s BSS (SBSS) and other BSS algorithms. The results demonstrate proposed algorithm efficiency in real time blind extraction of both super-Gaussian and sub-Gaussian signals from their mixtures.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: An automatic artifact extraction system is proposed based on a hybridization of Stone’s BSS and genetic algorithm. This hybridization is called evolutionary Stone’s BSS algorithm (ESBSS). Original Stone’s BSS used short- and long-term half-life parameters as constant values, and the changes in these parameters will be affecting directly the separated signals; also there is no way to determine the best parameters. The genetic algorithm is a suitable technique to overcome this problem by finding randomly the optimum half-life parameters in Stone’s BSS. The proposed system is used to extract automatically the common artifacts such as ocular and heart beat artifacts from EEG mixtures without prejudice to the data; also there is no notch filter used in the proposed system in order not to lose any useful information.
    Mathematical Problems in Engineering 08/2014; 2014. DOI:10.1155/2014/324750 · 1.08 Impact Factor


Available from