Comparative analysis of time frequency representations for discrimination of epileptic activity in EEG signals

Epilepsy is a brain pathology that affects approximately 40 million people in the world. The most utilized clinical test for epilepsy diagnose is the electroencephalogram (EEG). For this reason, nowadays are being developed multiple tools devised for automatic seizure detection on EEG signals. In this work, several approaches of TFR estimation for detection of epileptic events in EEG recordings are compared. Parametric (stochastic evolving and local estimation) TFR estimators as well as non-parametric (STFT, SPWV and CWT) are under study. Comparison is made according with the achieved performance using a recently proposed methodology for TFR based classification. Results show similar outcomings with all approaches for TFR estimation, achieving accuracy rates from 96 to 99%. Best performance was found for STFT and STTVAR approaches for TFR estimation.