Going live: Evaluating conventional and Riemannian classification techniques on the online recognition of mental workload
Keywords
Loading...
Authors
Issue Date
2022-06-28
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
The covert detection of mental workload levels (MWL) provides a way to adapt task
contexts to avoid mental overloads in critical job situations. Brain-computer interface
(BCI) systems can perform this task in controlled experiment settings, but it is
unclear if existing approaches are suitable for real-life situations. In this study, I will
compare two conventional classifiers, linear discriminant analysis (LDA) and support
vector machine (SVM), with recently developed Riemannian classifiers, minimum
distance to Riemmanian mean (MDRM) and tangent space LDA (TSLDA). The
study uses a novel experiment design in which the participant has free choice of the
task condition to perform. The results are evaluated on the basis of five ecological
criteria to measure BCI performance toward real-world use. MDRMand TSLDA significantly
outperform the conventional approaches in accuracy, stability and speed.
This study provides a bridge between controlled experiment settings and the real
world. Future research directions point towards refinement of experiment design
and ecological evaluation of other promising classifiers.
Description
Citation
Faculty
Faculteit der Sociale Wetenschappen