Epilepsy is one of the most common neurological disorders, affecting around 1 in 200 of the population. However, identifying epilepsy can be difficult because seizures tend to be relatively infrequent events and an electroencephalogram (EEG) does not always show abnormalities. The aim of this project is to develop a new methods that could improve the diagnosis of epilepsy, leading to earlier treatment and to a better quality of life for epileptic patients. The above methods must be composed with a flexible hardware development in order to discriminate noise and bad signals from correct EEG, MEG (Magnetoencephalogram), Eye Image recognition, Somnography and DTI (Diffusion Tensor Imaging). Even if there are EEG signal classifiers, it is suitable to perform a correct signal processing according to particular clinical reference, that is, it is difficult to have a classifier for all circumstances but it is possible to adapt EEG processing on current patient. Preliminary results are described for processing biomedical signals, namely EEG signals, in order to train the adaptive filtering in recognizing and choosing correct frequencies at which it is possible to reduce noise.