Describir: Automatic Detection of Paroxysms in EEG Signals Using Morphological Descriptors and Artificial Neural Networks