Development of EOG-Based Human Computer Interface (HCI) System Using Piecewise Linear Approximation (PLA) and Support Vector Regression (SVR)

Electrooculogram (EOG)-based human-computer interfaces (HCIs) are widely researched and considered to be a good HCI option for disabled people. However, conventional systems can only detect eye direction or blinking action. In this paper, we developed a bio-signal-based HCI that can quantitatively estimate the horizontal position of eyeball. A designed bio-signal acquisition system can measure EOG and temporalis electromyogram (EMG) signals simultaneously without additional electrodes. For real-time processing for practical application, modified sliding window algorithms are designed and applied for piecewise linear approximation (PLA). To find the eyeball position, support vector regression (SVR) is applied as a curve-fitting model. The average tracking error for target circle with a diameter of 3 cm showed only 1.4 cm difference on the screen with a width of 51 cm. A developed HCI system can perform operations similar to dragging and dropping used in a mouse interface in less than 5 s with only eyeball movement and bite action. Compare to conventional EOG-based HCI that detects the position of the eyeball only in 0 and 1 levels, a developed system can continuously track the eyeball position in less than 0.2 s. In addition, compared to conventional EOG-based HCI, the reduced number of electrodes can enhance the interface usability.