A method for building a real-time cluster-based continuous mental workload scale

Abstract

Mental workload is known to significantly influence an operator's task performance. Extreme levels of mental workload can lead to operator monotony, low performance and operation errors. Measurement of mental workload is thus very important–especially real-time measurement that involves the representation of continuous real numbers. Existing mental workload measures include human subjects’ self-reporting, task performance and psychophysiological signals. Subjects’ self-reporting and task performance measures are used posterior–i.e. after the task is performed–and thus they cannot be used for real-time measurement of mental workload. Measures based on psychophysiological signals are suitable for real-time measurement and currently are usually represented as a set of discrete numbers–so-called levels. Mental workload is essentially a quantity of continuous numbers and defining it as a set of discrete levels can introduce unnecessary constraints on the accurate understanding of mental workload. In this paper, a novel methodology is proposed for constructing a measure of mental workload with a continuous real number representation. This methodology is based on the view that mental workload is task dependent and its quantitative representation of high or low mental workload should be dependent on training data. For the purpose of illustrating and validating this methodology, the electrocardiogram (ECG) signal was used. The proposed ECG-based measure was compared with the Rating Scale Mental Effort (RSME) method in a driving application. The result of the validation has shown that the proposed method is in good agreement with the RSME method; however, it is known that RSME cannot be done in real time and with a representation of a set of levels.