Design and implementation of an eye gesture perception system based on electrooculography. (January 2018)

• Record Type: Journal Article
• Authors: Lv, Zhao; Zhang, Chao; Zhou, Bangyan; Gao, Xiangping; Wu, Xiaopei
• Is Part Of: Expert systems with applications. Volume 91(2018)
• Page Start: 310
• Page End: 321
• Publication Date: 2018-01
• Subjects: Electrooculography (EOG) -- Eye gesture -- Unit saccadic signals -- Common spatial pattern (CSP) -- Joint approximate diagonalization -- Support vector machine (SVM); 00-00 -- 99-00; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2017.09.017
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: We present an eye gesture perception system based on electrooculography. Spatial eye movement features are extracted by Common Spatial Pattern model. A Saccade Activity Detection method is designed to achieve online function. Results reveal reliability and validity of the proposed system. This system can be used as an effective complementation of video-based method. Abstract: People with motor diseases have suffered from deprivation of both verbal and non-verbal communication abilities. Fortunately, some of them still retain coordination of brain and eye-motor. To establish a stable communication way for these disabled people, this paper presents an eye gesture perception system based on Electrooculography (EOG). In order to implement a high-accuracy of unit saccadic EOG signals recognition, we propose a new feature extraction algorithm based on Common Spatial Pattern (CSP). We first establish a CSP spatial filter bank corresponding to 8 saccadic tasks (i.e., up, down, left, right, right-up, left-up, right-down, and left-down), then use it to linearly project raw EOG signals and treat the outputs as feature parameters. Furthermore, eye gestures recognition has been carried out by identifying and merging unit saccadic segments in terms of pre-defined time sequences. Experiential results over 10 subjects show that the recognition precision of unit saccadic EOG and eye gesture are 96.8% and 95.0% respectively, which reveal the proposed system has a good performance of eye gestures perception.