Wearable EOG Goggles: Eye-Based Interaction in Everyday Environments

Comments:

Paul’s Blog
Kevin’s Blog


Summary:

This paper presents an eye tracker which uses Electrooculography (EOG) for eye motion sensing and context recognition. The use of eye movement gestures has advantages of privacy, and is a natural indicator of attention and intention. Furthermore it does not alter the temporal sequence actions by requiring a dwell time to make selections. The EOG system has advantages over the video based eye trackers which are bulky, obtrusive and not geared for online embedded processing which is a critical necessity for eye movement based communication.

EOG was shown to be adequate to read eye movements robustly in everyday situations with a mobile setup. This was first tested with a commercially available system, after which a purpose designed one was constructed. Unlike previous studies which focussed on using eye movements for object selection, the present one was concerned with fast interaction. EOG recorded Saccades, fixations and blinks have been reliably used for robot control from fixed positions.

The EOG goggles designed and used in this project are a miniaturized highly wearable and self contained stand alone set up with embedded recording. In order to compensate for physical activity an ambient light meter and an accelerometer were incorporated into the frame. EOG measures the electric potential field surrounding the eye. Any movements of the globe, which itself is a dipole, result in a change in the surrounding potential field which is translated into a directional movement signal by the system.

Blinks are detected and removed in this particular application but can be used for selection etc. Consecutive Saccades in the horizontal and vertical planes are recorded. A Continuous Wavelet Transform - Saccade Detection algorithm is used for Saccade detection. Detected Saccades directions are mapped into “L”, “R”, “U” and “D”. Diagonals are determined from simultaneous combination of these and mapped into “1”, “3”, “7” and “9”. Recognition of several consecutive gestures is done through string matching.

The results for performing a test structured in the form of a game of increasing difficulty showed that 30% of the subjects had difficulty maintaining concentration, but overall the system outperformed speech and hand gestures.


Discussion:

The EOG system presented as a more compact and portable alternative to more traditional eye tracking devices has tremendous utility. Additionally, the use of the naturalistic modality of eye gestures is interesting.
However, without extensive practice the performance of unnatural eye gestures must be very distracting as it is using a modality which has a very high attentional priority in a very non natural manner. Repeatedly performed forced Saccades, independent of objects in the visual field would be potentially disturbing.
The use of this system in a mobile setting makes no mention of OKN or how it would be filtered out. There is some mention of an accelerometer but no indication as to how it is implemented. Are eye gestures only possible while stationary? Unlike video based eye trackers where fixation positions, or dwells are recorded in addition to Saccades no mention is made of how spontaneous Saccades would be separated from gestures.
The exact nature of the test is not identified, and the basis for concluding the superior performance of eye gestures in comparison to speech and hand gestures needs explanation.
This system would be better suited to non dynamic environments where spontaneous eye movements and the demands on the operators attention can be predictably controlled.