PROJECT TITLE :

Sparse EEG Source Localization Using Bernoulli Laplacian Priors

ABSTRACT:

Supply localization in electroencephalography has received an increasing amount of interest in the last decade. Solving the underlying ill-posed inverse drawback sometimes needs choosing an acceptable regularization. The usual $ell _2$ norm has been thought of and provides solutions with low computational complexity. However, in several situations, realistic brain activity is believed to be targeted in a few focal areas. In these cases, the $ell _two$ norm is known to overestimate the activated spatial areas. One resolution to the present problem is to market sparse solutions for instance based mostly on the $ell _one$ norm that are simple to handle with optimization techniques. In this paper, we have a tendency to contemplate the utilization of an $ell _0 + ell _one$ norm to enforce sparse source activity (by ensuring the answer has few nonzero elements) while regularizing the nonzero amplitudes of the answer. Additional exactly, the $ell _0$ pseudonorm handles the position of the nonzero components whereas the $ell _1$ norm constrains the values of their amplitudes. We have a tendency to use a Bernoulli–Laplace previous to introduce this combined $ell _0 + ell _one$ norm in a very Bayesian framework. The proposed Bayesian model is shown to favor sparsity whereas jointly estimating the model hyperparameters using a Markov chain Monte Carlo sampling technique. We tend to apply the model to each simulated and real EEG data, showing that the proposed technique provides better results than the $ell _two$ and $ell _1$  norms regularizations in the presenc- of pointwise sources. A comparison with a recent technique based on multiple sparse priors is additionally conducted.