Related Experiment Video
Updated: Dec 7, 2025
06:58
An Application for Pairing with Wearable Devices to Monitor Personal Health Status
Published on: February 3, 2022
3.2K
Using Machine Learning and Smartphone and Smartwatch Data to Detect Emotional States and Transitions: Exploratory
Madeena Sultana1,2, Majed Al-Jefri1,3, Joon Lee1,2,4
1Data Intelligence for Health Lab, Cumming School of Medicine, University of Calgary, Calgary, AB, Canada.
JMIR Mhealth and Uhealth
|September 29, 2020
Summary
This study shows that daily context, like location and phone use, can predict emotional states and transitions using smartphone data. This automated detection offers a promising alternative to self-reports for understanding well-being.
Area of Science:
- Computational Social Science
- Affective Computing
- Digital Phenotyping
Background:
- Daily emotional state is crucial for health, but self-reports are often inconvenient and incomplete.
- Automated detection of emotional states and transitions could offer a practical solution.
- The link between emotional changes and everyday context remains under-explored.
Purpose of the Study:
- To investigate the relationship between contextual information and emotional states/transitions.
- To evaluate the feasibility of detecting emotional states and transitions from daily contextual data using machine learning (ML).
Main Methods:
- Utilized smartphone and smartwatch sensor data (accelerometer, GPS, microphone, etc.) from 18 individuals.
- Mapped 49 self-reported emotions to the pleasure-arousal-dominance model, defining 6 emotional states.
- Developed general and personalized ML models for 5-minute interval emotion state and transition detection.
Main Results:
- Achieved high accuracy in personalized emotional state detection (avg. 96.33% AUROC) and transition detection (avg. 88.73% AUROC).
- Spatiotemporal context, phone state, and motion data were key predictors.
- Lifestyle significantly impacts emotion predictability.
Conclusions:
- Demonstrated a strong association between daily context and emotional states/transitions.
- Confirmed the feasibility of detecting emotional states and transitions using ubiquitous sensor data from wearables and smartphones.
