Automatic removal of EOG artifacts using SOBI algorithm combined with intelligent source identification technique

Pham Van, Thanh and Dang Nhu, Dinh and Nguyen Duc, Anh and Nguyen Tien, Anh and Chu Duc, Hoang and Tran, Duc Tan (2017) Automatic removal of EOG artifacts using SOBI algorithm combined with intelligent source identification technique. In: International Conference on Advanced Technologies for Communications, Oct., 18-20, 2017, Quy Nhon, Viet Nam.

PDF Download (514kB)

Abstract

Electrooculography (EOG) artifacts, generated by winking or other eye’s movements, should be eliminated because they are the cause of the wrong decision in analysis the Electroencephalography (EEG) data, especially in the diagnosis of epilepsy. One of the efficient methods for signal separation is the Second order blind identification (SOBI), a blind source separation technique. In most cases, the activities of the two eyes are the same, and SOBI identify that there is only one source of artifact. However, in some cases, the activities of the two eyes are different, and SOBI identify that there are two different sources of artifacts [1]. The problem is that SOBI cannot provide the information about the order of sources. It means that, it cannot point out how many sources of EOG. It would lead to the wrong decision in EEG analysis. To solve this current limitation, in this paper, we propose an effective method to remove EOG from EEG using SOBI combined with intelligent source identification technique. The proposed method was evaluated carefully using experimental data. It determined successfully the number of EOG sources and removes these artifacts more accurately and efficiently.

Item Type:	Conference or Workshop Item (Lecture)
Subjects:	Electronics and Communications > Electronics and Computer Engineering
Divisions:	Faculty of Electronics and Telecommunications (FET)
Depositing User:	Assoc/Prof Duc Tan Tran
Date Deposited:	29 Oct 2017 07:23
Last Modified:	10 Jan 2018 07:57
URI:	http://eprints.uet.vnu.edu.vn/eprints/id/eprint/2593

Actions (login required)

View Item