Calibrated photoplethysmography-derived heart rate variability approximates electrocardiographic gold-standard performance for stress-state classification
Om Chhaya, Viswanath Ananth
· 2026
Summary
WESAD (15 subjects): time/frequency/nonlinear HRV features. ECG features → 80% acc, ROC AUC 0.85; linear calibration mapping PPG HRV features into ECG space raised PPG ROC AUC 0.79 → 0.85. LSTM/CNN did not beat features.
Key facts
- Form Factor
- wrist
Reported benchmarks
Mental workload / stress
WESAD · AUROC ↑
#4 of 5
0.85
binary stress; classical ML on HRV features; PPG features linearly calibrated to ECG space (uncalibrated PPG 0.79; ECG 0.85); 15 subjects
Updated: Oct 5, 2026