• Rivero Cebrián, Daniel
• Dorado de la Calle, Julián
• Rabuñal Dopico, Juan Ramón
• Pazos Sierra, Alejandro

Biomedical signal processing is one of the research fields that has received more research in the recent years or decades. Inside it, signal classification has shown to be one of the most important aspects. One of the most used tools for doing this analysis are Artificial Neural Networks (ANNs), which have proven their utility in modeling almost any input/output system. However, their application is not easy, because it involves some design and training stages in which the expert has to do much effort to develop a good network, which is even harder when working with time series, in which recurrent networks are needed. This paper describes a new technique for automatically developing Recurrent ANNs (RANNs) for signal processing, in which the expert does not have to take part on their development. These networks are obtained by means of Evolutionary Computacion (EC) tools, and are applied to the classification of electroencephalogram (EEGs) signals in epileptic patients. The objective is to discriminate those EEG signals in which an epileptic patient is having a seizure.