Investigating the role of electroencephalogram based rhythmic bands: applications in imagined speech classification and epileptic seizure severity detection

The objective of this paper is to explore the significance of rhythmic bands for imagined speech classification and the severity detection of epileptic seizures from electroencephalogram (EEG) signals. Non-invasive decoding of imagined speech (words or sentences imagined in the brain) finds important applications in patients with lock-in syndrome and other neurological disorders. In the present work, the rhythmic analysis of EEG utterances is carried out by filtering each channel into five frequency bands: Alpha (8–12 Hz), Beta (12–35 Hz), Theta (4–8 Hz), Delta (0.5–4 Hz), and Gamma ( \(> 35 Hz\) $>35Hz$ ). From the comparative sub-band analysis carried out in the open-access Correto database, the Alpha rhythmic bands are confirmed to have more vowel information compared to other subbands. In the context of classifying the severity level of EEG signals from epileptic patients, the Gamma band yielded the best performance.