Abstract :
EEG events are widely used to diagnose patients who suffer from different diseases, including epilepsy. The EEG during a seizure exhibits characteristic temporal and spectral properties, depending upon the seizure type and its cause. Identifying an EEG with an ictal event of this nature can help to support diagnosis and may also be used to classify the type of seizure. From this work, based on time-frequency analysis pre-processing of EEG seizures, we obtained some good results about the best resolution of frequency changes for feature extraction used for neural net input. Together with the other features (from the same data mining), the system performs a neural net-based and knowledge-based detection. There has been no such method reported previously in the literature about how to determine a signature for an EEG event
