Information regarding the state of Warfighters (e.g., workload, fatigue, distraction) can greatly impact outcomes in training contexts and the likelihood of mission success in operational environments. Harnessing neural data collected via functional near-infrared spectroscopy (fNIRS) can be a game-changing tool for assessing individual and team states. However, most neural measurement devices are plagued by barriers of cost, size, portability, and ease of use by non-experts. The purpose of this research was to assess whether low-cost, LED-based fNIRS devices can appropriately classify operator workload states in varied task difficulty conditions. As such, a lightweight, easy-to-adopt fNIRS system was developed that translates raw neural data into estimates of functional states by providing access to cerebral oxygenation data from the prefrontal cortex (PFC). These data can indicate a variety of executive functioning activities relevant to operators, including attentional focus, inhibitory control, and stress control, among others (Fuster, 2015). Behavioral and physiological task data, and self-report data (NASA Task Load Index) were collected, with participants engaging in an updated version of the synthetic work environment (SynWin; Elsmore, 1994) called Aviator SynWin. Aviator SynWin, like its predecessor, is a computer-based environment that requires participants to simultaneously perform four unrelated tasks. Additionally, participants completed the n-back task (Kirchner, 1958), the Continuous Performance Test (CPT; Rosvold et al., 1956), and the Flanker Task (Eriksen, 1974) to provide calibration data to train models of operator state such as workload. Data collection is ongoing (80% complete) but will conclude with ample time to execute analyses before paper submission, including training individualized models of workload classification and assessing correlations between performance and physiology. We hypothesize that our models will classify higher workload states during more difficult task conditions and lower workload states during less difficult task conditions.
Keywords
MACHINE LEARNING;PHYSIOLOGICAL
Additional Keywords
fNIRS, cognitive workload