Authors:
Jean-François Gagnon
1
;
Olivier Gagnon
2
;
Daniel Lafond
1
;
Mark Parent
2
and
Sébastien Tremblay
2
Affiliations:
1
Thales Research and Technology, Canada
;
2
Université Laval, Canada
Keyword(s):
Operator Functional State, Psychophysiological Modeling, Machine Learning, Data Processing.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Artificial Intelligence
;
Biomedical Engineering
;
Biomedical Signal Processing
;
Biosignal Acquisition, Analysis and Processing
;
Data Manipulation
;
Health Engineering and Technology Applications
;
Human-Computer Interaction
;
Methodologies and Methods
;
Neurocomputing
;
Neurotechnology, Electronics and Informatics
;
Pattern Recognition
;
Physiological Computing in Mobile Devices
;
Physiological Computing Systems
;
Sensor Networks
;
Soft Computing
;
Software Engineering
Abstract:
This paper addresses critical issues and reports key findings with regards to the development of participant-generic operator functional state (OFS) models in the context of cognitive work. Conceptually, this research is concerned with the nature of the relationship between the physiological state of individuals and human performance. Participants were physiologically monitored (cardiac, respiratory, and eye activity) during the execution a set of two cognitive tasks – n-back and visual search – for which there were two levels of difficulty. Levels of difficulty were associated with levels of mental workload. Performance on the tasks was also monitored and linked with OFS. Modeling of the relationship between physiological state and OFS involved systematic manipulation of three parameters: (1) size of smoothing window for performance, (2) performance decrement threshold for labelling functional and sub-functional states, and (3) the mode of classification being either prospective or
descriptive. Modeling was performed using two types of classifiers. Results show that (1) models that use bio-behavioral data were capable of classifying performance on new participant data above chance, (2) levels of mental workload were better classified than OFS, (3) size of smoothing window had a significant impact on classifier performance, and (4) size of smoothing window, threshold values, and classifier type had a significant impact on sensitivity and specificity. Implications for the use of OFS models in operational contexts are discussed.
(More)