Unlocking Possiblities while Preserving Performance: Putting the ;;Interface;; back in Brain-Computer Interface.

Brain-computer Interface;Assistive Technology;Input Emulation;Biomedical Engineering;Engineering;Biomedical Engineering

Thompson, David EdwardSyed, Zeeshan ;

University of Michigan

关键词: Brain-computer Interface;    Assistive Technology;    Input Emulation;    Biomedical Engineering;    Engineering;    Biomedical Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/94006/dthomp_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Brain-computer interface (BCI) technology offers the hope of communication and control for people with the most severe motor impairments.Surveys of user populations indicate that users are interested in BCIs for a variety of tasks.Thus, an eventual goal for the BCI field should be flexible interfaces usable for multiple purposes.Yet these interfaces must not sacrifice performance for the sake of flexibility – BCI performance is already poor compared to existing assistive technology (AT).To succeed in the highly competitive market of AT, BCIs must do several jobs, and do them well.This dissertation presents the design and testing of the world;;s first plug-and-play BCI, capable of interfacing with many existing AT devices in addition to most personal computers.Both a communication task and an environmental control task were tested using the system.The communication test indicates that the plug-and-play BCI can be used to operate AT communication devices with minimal performance cost (95% confidence bounds indicate an accuracy difference smaller than 3.5 percentage points).The control test indicates that the plug-and-play BCI can be used to operate a wheelchair seating system with small performance costs (accuracy difference less than 9 percentage points).The dissertation also includes insights into the issue of performance measurement in the BCI field.A review and critique of existing BCI performance metrics was performed, including a comparison based on data from earlier experiments.Based on this comparison, Information Transfer Rate and BCI-Utility are suggested for broad use in the BCI field.Finally, a novel method of accuracy estimation, using classifier-based latency estimation (CBLE) is developed and presented.The accuracy estimates from the new method are significantly more correlated with daily accuracy than estimates based on either training accuracy (r = 0.64 vs. 0.2, p < 0.05) or accuracy on a small dataset (r = 0.74 vs. 0.16, p < 0.05).As BCI experiments often have relatively small datasets, the method has potential to increase the power of many experiments across the field.In addition to addressing broadly applicable performance measurement issues, the dissertation thus increases the available options for BCI users while preserving performance.