Dynamic Stopping for cVEP-based Brain-Computer Interfaces (BCI’s)
Keywords
Loading...
Authors
Issue Date
2022-04-27
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Objective. In Brain-Computer Interfaces (BCI’s), the most important part
of the process is to accurately and quickly classify the brain signals of a user.
This helps the system to provide a quick and accurate output and improves
user experience. An important part of this classification is deciding when to
stop gathering data and producing the output. The literature suggests that
’dynamic stopping’ as opposed to ’static stopping’ methods, have the best
performance. A lot of dynamic stopping methods have been created in the
literature, but it is unclear what the benefits of these existing methods are.
The goal of this research is to get an indication of their relative performance
and benefits.
Approach. The performance of multiple stopping methods was calculated
using offline analysis on a large data set. Furthermore, these methods were
all optimized by tweaking their variables to get the optimal performances.
This performance was calculated in typed Symbols per minute (SPM) because
they were applied to a BCI speller framework.
Main results. The results clearly show that using dynamic stopping methods
has a positive effect on the performance of a BCI as opposed to using a
static stopping method. Every subject would get the best BCI performance
by using one of the dynamic methods. Those dynamic methods performed
comparable to each other, but have high variance depending on the subject.
Optimizing the methods did not create large improvements in performance
compared to the training necessary to facilitate them. Only one of the dynamic
methods tested did not need long training sessions to facilitate high
performance.
Significance. The results in this research shows how stopping methods used
in cVEP-BCI compare to each other and highlight their individual benefits.
It also allows stopping methods designed in future research to be easily compared
to its competitors. The code created for this research also facilitates
this easy implementation and comparison.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen