Cognitive load detection from wrist-band sensors

摘要: In this work, we use machine learning (ML) to detect the cognitive load of a user based on sensor data from smart wrist-band, sampled during 30 seconds. The is provided by challenge at UbiTtention 2020 workshop UbiComp 2020; in paper describe UW's participation (team Lynx). defining characteristic our approach custom features that extract time series. While do not have any labeled instances for test users, fact multiple series each user, allows us measure how much individual deviate user's average. We combine extracted information with other series' literature. further feature selection Gini impurity and state-of-the-art techniques training ML models such as Logistic Regression, (Boosted) Decision Trees, Random Forests, Support Vector Machines, yielding~63% accuracy 6-fold cross-validation.