Classification of Seizure EEGs Based on Short-Time Fourier Transform and Hidden Markov Model

Epilepsy is a kind of disorder that has affected many people in the world. Electroencephalogram (EEG) is an effective tool in the diagnosis and treatment of epilepsy. The classification of EEG signals from different seizure stages is of great interest in this field. This paper proposes a seizure EEG classification method based on Short-Time Fourier Transform (STFT) and Hidden Markov Model (HMM). We construct feature sequences by STFT, and then 50% of the sequences are used to train HMMs. Finally, the other sequences are used to evaluate the HMMs. Experiments conducted on the dataset from University of Bonn are provided, with the accuracy for set D and set E reaching 97.18%, and the sensitivity and specificity reaching 98.54% and 95.82% respectively.