EEG-based classification of visual and auditory monitoring tasks

Abstract

Using EEG signals for mental workload detection has received particular attention in passive BCI research aimed at increasing safety and performance in high-risk and safety-critical occupations, like pilots and air traffic controllers. Along with detecting the level of mental workload, it has been suggested that being able to automatically detect the type of mental workload (e.g., auditory, visual, motor, cognitive) would also be useful. In this study, we developed a novel experimental protocol in which subjects performed a task involving one of two different types of mental workload (specifically, Auditory and Visual), each under two different levels of task demand (Easy and Difficult). The tasks were designed to be nearly identical in terms of visual and auditory stimuli, and differed only in the type of stimuli the subject was monitoring/attending to. EEG power spectral features were extracted and used to train linear discriminant classifiers. Preliminary results on three subjects suggested that the Auditory and Visual tasks could be distinguished from one another, and individually from a Baseline condition (which also contained nearly identical stimuli that the subject did not need to attend to), with accuracy significantly exceeding chance. This was true when classification was done within a workload level, and when data from the two workload levels was combined. Though further investigation is required, these preliminary results are promising, and suggest the feasibility of a passive BCI for detecting both type and level of mental workload.