We collected electroencephalographic (EEG) data from 16 subjects while they performed a mental arithmetic task at five different levels of difficulty. A classifier was trained to discriminate between three conditions: relaxed, low workload and high workload, using spectral features of the EEG. We obtained an average classification accuracy of 62%. A continuous workload index was obtained by low-pass filtering the classifier’s output. The correlation coefficient between the resulting workload index and the difficulty level of the task was 0.6 on average.
Get full access to this article
View all access options for this article.
References
1.
GevinsA.SmithM. (2003). Neurophysiological measures of cognitive workload during human-computer interaction. Theoretical Issues in Ergonomics Science4(1), 113–131
2.
HankeM.HalchenkoY.O.HaxbyJ.V.PollmannS. (2010). Statistical Learning Analysis in Neuroscience: Aiming for Transparency. Frontiers in Neuroscience, 4.
3.
KohlmorgenJ.DornhegeG.BraunM.BlankertzB.MüllerK.CurioG.HagemannK.BrunsA.SchraufM.KincsesW. (2007). Improving human performance in a real operating environment through real-time mental workload detection. Toward Brain-Computer Interfacing, pp. 409–422
4.
ParasuramanR.MoulouaM.MolloyR. (1996). Effects of adaptive task allocation on monitoring of automated systems. Human Factors, 38, 665–679
5.
PutzeF.JarvisJ.SchultzT. (2010). Multimodal Recognition of Cognitive Workload for Multitasking in the Car. International Conference on Pattern Recognition 2010
6.
WienerE.L. (1985). Beyond the sterile cockpit. Human Factors, 27, 75–90
