The design of a hybrid text-entry system using a self-paced brain-computer interface and an eye-tracker

• Main Author: Yong, Xinyi
• Language: English
• Published: University of British Columbia 2012
• Online Access: http://hdl.handle.net/2429/42786

This thesis proposes a hybrid brain-computer interface (BCI) system that consists of a self-paced BCI (SBCI) and an eye-tracker to operate a virtual keyboard. The design aims to overcome the difficulties encountered by individuals with disabilities that prohibit them from using a standard keyboard and a mouse. To make a text-entry of a letter/word, the user of the proposed system should gaze at the target for at least a specific period of time (called the dwell time) and then attempt a hand extension movement that activates the SBCI. Although the SBCI is available for use at any time, a built-in sleep mode is activated when the user is not looking at a letter/word or when the user gazes at a letter/word for less than the dwell time. Such a design has the advantage of greatly minimizing the false positive (FP) outcomes compared to state-of-the-art SBCIs. While operating a pure or a hybrid BCI, artefacts such as ocular and muscle artefacts frequently contaminate the EEG signals and subsequently the BCI's performance is degraded. To overcome this problem, the thesis proposes two algorithms for detecting and removing various types of artefacts. Both algorithms are fully automatic and do not use additional electrooculogram, electromyogram or frontal/temporal EEG channels. They also allow real-time processing. The proposed artefact removal algorithm does not require long data segments and reduces the signal distortion. When the artefact detection and removal algorithms are applied to our hybrid BCI, we demonstrate that for dwell time threshold of 0.0s, the number of FPs/minute is two and the true positive rate (TPR) is 44.7%, which is more than 15.0% higher compared to other state-of-the-art artefact handling methods. As the dwell time increases to 1.0s, the TPR increases to 73.1%. Finally, an algorithm that adaptively updates the classifier of the hybrid BCI during online operation is proposed. This allows the system to adapt to changes in the properties of EEG signals and improves its performance. We show that in our application, the proposed algorithm successfully reduces the number of FPs generated per minute. It also outperforms other state-of-the-art adaptive classifiers used in this study.