Key result
A Bayesian deep learning model achieved a peak F1 score of 0.7 for classifying emotional valence using unimodal heartbeat time series collected by a consumer fitness tracker.
Why the study?
Real-world physiology-based emotion detection requires using ubiquitous PPG data from consumer fitness trackers rather than ECG, alongside uncertainty measures for healthcare settings.
Does a Bayesian deep learning model using unimodal PPG data from a consumer fitness tracker accurately predict emotional valence?
Does a Bayesian deep learning model using unimodal PPG data from a consumer fitness tracker accurately predict emotional valence?
Absolute Event Rate: 0.7% vs 0.57%
A Bayesian deep learning model can predict emotional valence from consumer fitness tracker PPG data with an F1 score of 0.7, demonstrating the potential for real-world physiology-based emotion detection.
May support wearable-based emotion monitoring; leaves open validation in larger cohorts before any clinical application.
Automatic detection of emotion has the potential to revolutionize mental health and wellbeing. Recent work has been successful in predicting affect from unimodal electrocardiogram (ECG) data. However, to be immediately relevant for real-world applications, physiology-based emotion detection must make use of ubiquitous photoplethysmogram (PPG) data collected by affordable consumer fitness trackers. Additionally, applications of emotion detection in healthcare settings will require some measure of uncertainty over model predictions. We present here a Bayesian deep learning model for end-to-end classification of emotional valence, using only the unimodal heartbeat time series collected by a consumer fitness tracker (Garmin Vivosmart 3). We collected a new dataset for this task, and report a peak F1 score of 0.7. This demonstrates a practical relevance of physiology-based emotion detection `in the wild' today.
No takes yet. Share an insight, caveat, or question.
Harper et al. (2019) studied Emotion detection (n=17). Bayesian deep learning model using unimodal PPG data vs. Chance model was evaluated on Peak F1 score for binary classification of emotional valence. A Bayesian deep learning model achieved a peak F1 score of 0.7 for classifying emotional valence using unimodal heartbeat time series collected by a consumer fitness tracker.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: