Recognition of the States of Confusion during Assembly Work based on Electrodermal Activity

We generated an 11-dimensional feature vector, including feature quantities in the time domain based on electrodermal activity signals acquired from 16 workers to detect the degree of confusion of the worker in the assembly work and constructed classifier models. As a result of preliminary experiments and evaluations, the random forest model with a feature vector of a 5 s window size showed the highest macro-averaged F1-score of 0.657.