Epileptic seizures warning system

The present invention discloses a seizure warning system includes a preprocessing module, a feature extraction module, a classification module and the post-processing module; four modules sequentially connected; feature extraction module when using empirical mode decomposition algorithm to extract the frequency of each data segment frequency characteristics of the variance, frequency characteristics, and taking the data segment as the corresponding data segment feature value; classification module SVM algorithm determines the onset of a state converted to binary classification, low computational complexity, real-time, feature can be used to quickly identify changes in EEG signal data and identifies whether or seizures, seizure warning realized. The present invention is applicable to patients with epilepsy ECoG signals obtained high sensitivity, low error rate and a shorter time delay.