2019/07/16 by Ross J. Harper, Harper, Ross, Joshua Southern +1 · 1 citation
Medicine · Psychology · #FOS: Computer and information sciences #Heart Rate Variability and Autonomic Control #Human-Computer Interaction (cs.HC) #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Mental Health Research Topics
paper · pdf · doi:10.48550/arxiv.1907.07327
openalex publication_date 2019/07/16 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28
Automatic detection of emotion has the potential to revolutionize mental\nhealth and wellbeing. Recent work has been successful in predicting affect from\nunimodal electrocardiogram (ECG) data. However, to be immediately relevant for\nreal-world applications, physiology-based emotion detection must make use of\nubiquitous photoplethysmogram (PPG) data collected by affordable consumer\nfitness trackers. Additionally, applications of emotion detection in healthcare\nsettings will require some measure of uncertainty over model predictions. We\npresent here a Bayesian deep learning model for end-to-end classification of\nemotional valence, using only the unimodal heartbeat time series collected by a\nconsumer fitness tracker (Garmin V 'ivosmart 3). We collected a new dataset for\nthis task, and report a peak F1 score of 0.7. This demonstrates a practical\nrelevance of physiology-based emotion detection `in the wild' today.\n