Unobtrusive Smart-Watch Based Approach for Assessing Mental Workload

This study is a proof of concept for assessment of mental workload based on breathing pattern reconstructed from photoplethysmograph (PPG) signal. Mental workload is the amount of mental effort required while doing a task. The performance of an individual declines if mental workload is too high or too low; finally, resulting in mental fatigue or boredom. In this study, we have proposed a smart-watch based unobtrusive approach which can be used to monitor mental workload. Two sets of tasks are selected to impart low and high mental workload on an individual. The raw PPG data obtained from a smart watch is used to extract the breathing pattern while executing the tasks. Next a set of features are derived from these signals which are then analyzed for their predictability and correlation with mental workload. Classification is done on a set of selected features. Results indicate that we are able to achieve approximately 78% accuracy (using Hoeffding tree classifier) with a small subset of features for differentiating low and high mental workload. Paired t-Test shows other classifiers perform at par, which validates the importance of selected features. In addition, area under curve for receiver operating characteristics (ROC) and precision-recall curve (PRC) analysis shows selected Hoeffding tree classifier has better average performance and is also suitable for wearable devices.