Brain activity reconstruction from EEG data. Our work has been focused mainly on the statistical modeling of the brain sources in the brain, e.g. (Alecu, Voloshynovskiy and Pun, 2004). This has yielded a generic representation of the data under the form of Infinite Mixture of Gaussians (IMG) through the newly introduced Gaussian Transform of distributions, and the corresponding IMG estimation techniques (Alecu, Voloshynovskiy and Pun, 2005a, and 2005b). The IMG estimation procedures have then been adapted to the EEG inverse problem (Alecu et al, 2005), where they have produced the IMG EEG class of inversion methods. These are mainly robust focalization methods, using non-Gaussian data priors. The proposed algorithms also integrate results from our previous studies on the reconstruction bounds, and are able to incorporate iteratively other types of constraints such as MRI derived anatomic constraints.