hmilogo

Publication

Type of publication:Dissertation
TitleMaking brain-computer interfaces better: Improving usability through post-processing
Year published 2014
Bibtex cite IDplass2014making
KindPhD Thesis
LocationUniv. of Twente
Date 2014-11-21
AbstractBrain-computer interfaces (BCIs) allow you to control things directly with your mind. Unfortunately, such input devices based on observations of the body are plagued by noise, non-stationarities, and ambiguity. In the lab, we can protect systems somewhat from these influences, but in ‘the real world’, BCIs could use a little help. How important is good control anyway? How well can users even assess their level of control? Fourteen participants evaluated three sets of mental tasks each for five weeks. Most important to them was good task recognition and easy task execution. When people know the input they provide, they have a good perception of their level of control. Eighty-seven participants played a browser game with varying levels of control. The actual amount of control explained 72% of the control they thought they had. Post-processing is a simple solution to improve the recognition of brain signals and make it easier to provide. Post-processing changes the way de- tected brain signals are actually being used in an application. Although post- processing is standard practice with other inputs, this is not yet the case with BCIs. Of the more than 200 BCIs published about until 2006 only 15% used post-processing, according to an earlier literature study. A follow-up review shows that post-processing methods are still under-appreciated in BCI research, even though the improvements using these methods look very promising! To stimulate conscious use of and discussion about these post- processing methods, I provide a method overview with guidelines for appli- cation. At the same time, it is important to test these methods in practice. The goal of an experiment with eighteen participants was to reduce the nec- essary effort with post-processing. Although it did reduce the amount of active task execution time, this did not result in the expected reduction in perceived effort. Switching between the active and passive tasks cost more effort. This work confirms the importance of good control to the user and offers BCI researchers and developers a solution: post-processing. An overview and guidelines are provided to stimulate deliberate use and discussion. The research also shows how essential user tests are.
PublisherCentre for Telematics and Information Technology, University of Twente
Publisher LocationEnschede
ISSN1381-3617
ISBN978-90-365-3779-7
Number of pages173
Authors
D. Plass-Oude Bos 
Promotors
A. Nijholt 
Ass. Promotors
M. Poel 
Topics
Multimodal Interactions
Computational Intelligence
Human Factors
BiBTeX BiBTeX
Eprints[Eprints]
Official URLhttp://dx.doi.org/10.3990/1.9789036537797
Files
[PDF Download PDF document]

old Parlevink website   colophon   [Back] .