Real-time classification of EEG signals implemented on DSPIC for the diagnosis of epilepsy

In order to improve tools for understanding the complicated brain behavior of epilepsy disease, we analyze the ElectroEncephaloGram (EEG) signal which is a medical imaging technique for measuring the electrical activity of the brain. For the detection of the epileptic activity, several traditional methods are so tedious. In this context, we proposed an automated diagnostic system to examine cerebral epileptiform discharges. This dissertation divided into two steps: in the first step, we extracted features in time-domain features of the EEG signal, then we optimized features by the degree of characterization noted “J”, these attributes are ready to be integrated as input for the statistic classification method. In the second step, we implemented the results found previously on DSPIC in real time. In this paper, we used an EEG dataset collected from Bonn University open source database which contains 100 EEG signals for normal subjects and 100 for epileptic ones. The results show that the accuracy rate of statistic classification is 97.5% using optimized input features which are most performing than the results obtained by all extracted features with an accuracy rate equal 88%.