Cross-subject classification of mental fatigue by neurophysiological signals and ensemble deep belief networks
2017
Assessing mental fatigue (MF) by using neurophysiological signals is prospective for predicting instantaneous degradation of operator performance in safety-critical human-machine systems. Reliable MF classifier modeled and tested by the physiological data collected from different subjects is quite practical since such cross-subject paradigm avoids preparing comprehensive subject-specific training sets. This paper proposes a novel EEG-based cross-subject MF classifier, ensemble deep belief networks (EDBN), by exploiting advantages of the deep leaning principle for abstraction higher-level EEG representations. The EDBN framework builds two different DBNs for each training subject as the static feature abstractor and the adaptive estimator to track the novel physiological property in EEG abstractions of the testing subject. The temporal OFS is predicted by switching the ensembles of the two DBNs at each time step via a Gaussian-kernel based criterion. The competence of the EDBN is validated by examining the OFS classes defmed by the mental workload, mental fatigue, and the coupling effect of the two variables stimulated by the AutoCAMS platform. The comparison of the cross-subject OFS classification accuracy demonstrates the EDBN significantly outperforms several state-of-the-art classifiers.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
21
References
2
Citations
NaN
KQI