Recognising situation awareness associated with different workloads using EEG and eye-tracking features in air traffic control tasks

The rate of human errors would increase as air traffic control officers (ATCOs) lose situation awareness (SA), which could also be affected by their perceived workloads. Recognising ATCOs SA inadequacy is crucial for ensuring traffic safety. This paper aims to propose a two-phase analytical methodology for revealing SA-related neuro-physiological patterns and hierarchically recognising ATCOs SA loss with workload concerns using EEG and eye-tracking data. A simulated air traffic control (ATC) radar-monitoring experiment involving different task loads with SA-probe tests was first conducted to collect behavioural and physiological (EEG and eye-tracking) data simultaneously, and the NASA Task Load Index (NASA-TLX) scale was used to measure participants’ perceived workloads. In our two-phase methodology, behavioural data representing the task performance was analysed in Phase I using the Gaussian Mixture Model to determine sample’s SA, and the perceived workloads on samples were labelled using NASA-TLX scores. Subsequently, in order to achieve our purposes, the physiological data were annotated based on results from Phase I, the physiological feature base was extracted using a fast Fourier transformation and Hilbert transform in Phase II, and the linear discriminant analysis was then used to extract the core features as inputs to train multiple classifiers. Results showed that the neuro-physiological behaviours of SA loss during normal workloads differed from those in high workload situations. A leave-one-subject-out cross-validation was also performed, and the results demonstrated that the optimal performance was 76.1% for classifying high/low SA (1-Level classification) and 82.7% for recognising low SA associated with high workload (2-Level classification).