Automatic artefact compensation in EOG signals

Type: Semester thesis
Students: Martin Wirz, Philip Herter
Advisor: Andreas Bulling, Johannes Schumm
Project: Eye Movements and Visual Perception

The objective of this thesis was to implement and evaluate algorithms in order to be able to track a person’s gaze direction using a wearable EOG device. The focus was laid on noise, baseline drift and eye blink artifact removal. Algorithms for automatic detection and compensation of these artifacts were implemented and verified.

In-depth studies evoked that slow changing baseline drift is difficult to estimate in continuous EOG signals. To overcome this issue, a system was proposed which reduces baseline drift by segmentation of the signal into its movement characteristics with individual drift estimation on each segment. This results in a drift-reduced EOG signal.
Another topic investigated in this thesis is the calibration of an EOG system. Displaced electrodes around the eyes and different skin resistance between subjects requires a calibration for gaze estimation. An automatic calibration routine was proposed in order to retrieve a relation between EOG signal values and gaze angles. With the Artefact Compensation Function (ACF), a system was presented combining artefact removal, signal segmentation, drift reduction and calibration to obtain an improved EOG signal for higher level processing.

During the thesis, a GUI for signal processing was implemented (see Figure). It includes most of the evaluated algorithms as well as the possibility to observe their direct influence on the EOG signal.