Sensitive, Diagnostic and Multifaceted Mental Workload Classifier (PHYSIOPRINT)

摘要

Mental workload is difficult to quantify because it results from an interplay of the objective task load, ambient and internal distractions, capacity of mental resources, and strategy of their utilization. Furthermore, different types of mental resources are mobilized to a different degree in different tasks even if their perceived difficulty is the same. Thus, an ideal mental workload measure needs to quantify the degree of utilization of different mental resources in addition to providing a single global workload measure. Here we present a novel assessment tool (called PHYSIOPRINT) that derives workload measures in real time from multiple physiological signals (EEG, ECG, EOG, EMG). PHYSIOPRINT is modeled after the theoretical IMPRINT workload model developed by the US Army that recognizes seven different workload types: auditory, visual, cognitive, speech, tactile, fine motor and gross motor workload. Preliminary investigation on 25 healthy volunteers proved feasibility of the concept and defined the high level system architecture. The classifier was trained on the EEG and ECG data acquired during tasks chosen to represent the key anchors on the respective seven workload scales. The trained model was then validated on realistic driving simulator. The classification accuracy was 88.7 % for speech, 86.6 % for fine motor, 89.3 % for gross motor, 75.8 % for auditory, 76.7 % for visual, and 72.5 % for cognitive workload. By August of 2015, an extended validation of the model will be completed on over 100 volunteers in realistically simulated environments (driving and flight simulator), as well as in a real military-relevant environment (fully instrumented HMMWV).