The electrical dipole source localization in the 3-D brain structure is performed using a realistic head model based on an average human MRI data released by the Montreal Neurological Institute (MNI). The EEG data is preprocessed using integer spline wavelet prefiltering to enhance the visual inspection of the multichannel waveforms used for localization. The forward and inverse problems are solved by Boundary Element Method (BEM) and MUSIC (Multiple Signal Classification) scanning algorithms, respectively. The technique is tested on clinical cases with secondary epileptic seizures based on morphological lesions that could be depicted in the anatomical MRI examinations of the patients. The estimated locations of the generators of interictal epileptiform signals correspond well to the surrounding neural tissue of the lesions. The wavelet prefiltering facilitates the visual inspection and detection of the interictal epileptic spikes and the BEM based on a realistic head model gives an accurate anatomical localization for the focus. A focal source model is widely accepted in many cases of epilepsy and the MUSIC scanning seems to be very suitable for localizing these type of sources. The method is currently modified to handle the localization of cortical and distributed source models for the EEG which seem to be more appropriate for the event related potentials.