Calibration of time features and frequency features in the time-frequency domain for improved detection and classification of seizure in newborn EEG signals

Abstract

This paper presents new time-frequency features for seizure detection in newborn EEG signals. These features are obtained by calibrating relevant time features and frequency features in the joint time-frequency domain. The proposed features allow the possibility of improving the performance of the seizure detection and classification system based on multi-class SVM classifier.