Newborn’s EEG seizure detection using compact kernel time-frequency distributions and Doppler-lag domain features

In this paper, we deal with newborn’s EEG seizure detection. The presented approach is combined with previously proposed methods in literature in order to enhance detection performance. The features extracted from Doppler-lag domain are comparatively assessed using receiver operator characteristic (ROC) analysis measure. On the other hand, we combine further features that are obtained using time-frequency signal analysis tools. In particular, the recently proposed time-frequency distributions (TFDs) based on kernels with compact support (KCS) are employed and compared to the best-known quadratic representations. The obtained results show detection performance improvement by up to 1.5% compared to the other investigated methods.