- Andersen.pdf (3M)
Oslo and Akershus University College of Applied Sciences
Master in Universal Design of ICT
The number of consumer grade Electroencephalography (EEG) based brain-computer interfaces (BCI) on the market has increased considerably in recent years. New BCI technologies are emerging every year and the summary of the top BCI projects for the BCI award in 2013 (g.tec, 2017) states that “With so many new groups, ideas, and BCI directions, it can be difficult to identify the most impressive projects and trends. “ (Christoph Guger, 2014, s. 137). The Gartner Hype Cycle has listed “Brain-Computer Interface” at on the rise for the last two years (Chaffey, 2017). Most of these consumer grade devices use affect-detection for identifying emotions such as happy, angry, calm etc. (NeuroSky, 2017; Neurowear, 2017; Robertson, 2017) to improve or control their devices. Some devices also have support for mental commands using motor imagery and visual imagery (Emotiv Inc., 2017; Bobrov, et al., 2011). Former research on commercial grade headsets have mainly focused on general performance compared to the medical grade headsets (Bobrov, et al., 2011; Matthieu Duvinage, 2013) and/or testing performance using motor imagery mental commands (Cornelia Kranczioch, 2014), visual imagery mental commands (Bobrov, et al., 2011) or P300 (Matthieu Duvinage, 2013). Motor imagery, Visual imagery and P300 are all examples for ways to use an EEG based BCI to control a computer. To my knowledge, in this field any official universally designed standard for mental commands is still lacking, and I have not found any papers that compares motor imagery, visual imagery and/or P300 in order to check if either is better suited for universal design. Several companies are in the process of developing products that integrates consumer grade BCI into their products. Research has yet to investigate consumer grade BCI from a user centred design perspective. In this study I compared the real control and perceived control (confirmation bias) between motor and visual imagery with consumer grade headsets manufactured by Emotiv on 21 participants. The aim was to determine if either motor or visual imagery is less susceptible to confirmation bias, where the user think they are in control when they in reality are not. The results could be used to start working on a universally designed standard where motor or visual imagery might be used. I compared how much control participants reportedly felt they had over a videogame character (score of 1 to 10) while they played using an Emotiv EEG headset to move the character with motor imagery for 10 rounds and visual imagery for another 10 rounds. Unbeknownst to the test subjects, 5 of the rounds for motor and visual imagery were fake and they were merely watching a round played by someone else. Before the final test with 21 participants I did one pilot test with 6 participants and two pre-tests with 10 participants each to optimize the testing procedure before the final test. After each test I did a T-test on the control scores when in control and not in control for both motor and visual imagery. The T-tests comparing average control scores from the 21 participants in the final test resulted in a P-value of 0.70 for motor imagery and 0.53 for visual imagery. There was no significant difference (P < 0.05) between the control scores for when the participants were in control and not in control for both motor and visual imagery. The results from the final test was that 12 of 21 participants (57.1%) reported higher control scores when not in control for motor imagery rounds, and 12 of 21 participants (57.1%) reported higher control scores when not in control for visual imagery rounds. 8 of 21 participants (38%) reported higher control scores when in control for motor imagery rounds, and 8 of 21 participants (38%) reported higher control scores when in control for visual imagery rounds. One participant (4.7%) reported the same control score regardless of being in control or not for both motor and visual imagery. 5 of 21 participants (23.8%) reported higher control scores when in control for both motor and visual. 9 of 21 participants (42.8%) subjects reported higher control scores when not in control for both motor and visual. Three of 21 participants (14.2%) reported higher control scores when in control during motor imagery rounds, but not during visual imagery rounds. Three of 21 participants (14.2%) reported higher control scores when in control during visual imagery rounds, but not during motor imagery rounds. My findings in the final test suggest there is no significant difference between the control scores for when the participants were genuinely in control and when they were not in control. My findings also could not find a difference between motor and visual imagery. Most participants reported control scores that did not correlate with when they were genuinely in control. I was not able to reject my null hypothesis (H 0). More research is needed in order to come closer to a universally designed standard. A universally designed standard for EEG based BCI could benefit from being one where there is a correlation between perceived control and genuine control.
Permanent URL (for citation purposes)