A Multi-Sensor Approach to Automatically Recognize Breaks and Work Activities of Knowledge ...
summary
449 labeled activities from nine knowledge workers drive evaluation using smartwatch EDA and BVP, wrist accelerometer, phone logging, and RescueTime laptop traces, segmented into five minute windows and fed through imputation, resampling, feature selection and a gradient boosting classifier. A multi-device self-report strategy (participants prefer Devo) collects labels. Results report 94% F1 for work and 69% F1 for breaks, beating time-of-day and laptop-inactivity baselines by about 5–10 and 12–54 percentage points, and show breaks with higher physiological arousal, more phone use and less laptop activity.
AI-written summary · about
related talks
45mScience and Tech Backed Approach to Increase Productivity
06mCrowdAct: Achieving High-Quality Crowdsourced Datasets in Mobile Activity Recognition
18mAlways On(line)? User Experience of Smartwatches and their Role within Multi-Device Ecologies
06mUnderstanding User Contexts and Coping Strategies for Context-aware Phone Distraction Management ...
06mUnderstanding Physical Practices and the Role of Technology in Manual Self-Tracking
27mDevelopers Expose What's Breaking Productivity in Their Companies
01h 00mTake Control of the Data of You
01h 06mPersonal Knowledge Management Logseq Methods and Madness