Diagnosis of sleep disorders is still a challenging issue for a large number of nerve diseases. In this sense, EEG is a powerful tool due to its non-invasive and real-time catacteristics. This modality is being more and more used for diagnosis such as for epilepsy. It is also becoming widely used for many redictive, Preventive and Personalized Medicine (PPPM) applications.To understand sleep disorders, we propose a method to classify EEG signals in order to detect abnormal behaviours that could refect a specificmodification of the sleep pattern. Our method consists of extracting the characteristics based on temporal and spectral analyses with different descriptors. A classifcation is then performed based on these features. Validation on a public available database show promizing results withhigh accuracy levels.