Estimation of Glance from EEG using cursor control

Abstract

The variations in the electrooculogram (EOG) caused by eye motion are roughly proportional to the instantaneous horizontal and vertical glance angle. This linear correlation is exploited in systems using EOG to control software, and hardware such as artificial limbs, or wheelchairs. In these approaches, the drift in the electronics is commonly compensated by applying a high-pass filter. Consequently, the remaining EOG signal contains only blinks and rapid eye movement. However, repeating these eye gestures voluntarily is exhausting. This paper presents an algorithm that estimates the instantaneous glance of a subject from EEG signals. The subject is seated in front of a computer screen to control an application by glance. Because the visual field of interest, in this setting, is the limited area of the monitor, we can compensate the error in the glance estimate by detecting outliers. Because no high-pass filter is applied to the data, the user controls the applications by eye glance, which is comfortable and can be performed over extended periods of time. The numerical evaluation of the experiments with 12 volunteers, and video recordings of EOG controlled applications demonstrate the accuracy of our algorithm.