Traditional Signal Processing vs Deep Learning Algorithm - Real-World Driving Test Results
| Item | Value | Description |
|---|---|---|
| Test Date | 2026-01-05 | Real-world driving test date |
| Reference Device | Polar H10 | Chest-worn ECG Sensor (Heart Rate Measurement Standard) |
| Test Vehicle | Kia Niro Hybrid | Kia Motors Niro Hybrid |
| Driving Distance | 55.4 km | Real road driving test distance |
| Session | Algorithm | Time | Duration | Data Points |
|---|---|---|---|---|
| Test 1 | Deep Learning (rPPG-Net) | 12:40 ~ 13:10 | 27.5 min | 1,651 |
| Test 2 | Traditional Signal Processing | 14:44 ~ 15:43 | 58 min | 3,483 |
Real-world driving test results vs Polar H10 ECG reference
| Metric | Traditional | Deep Learning | Improvement | Winner |
|---|---|---|---|---|
| MAE (bpm) | 8.27 | 6.33 | +23.5% | DL |
| RMSE (bpm) | 9.90 | 7.68 | +22.4% | DL |
| Bias (bpm) | -6.23 | -3.53 | +43.3% | DL |
| Face Detection | 73.9% | 86.6% | +17.2% | DL |
| ±5 bpm Accuracy | 36.4% | 47.3% | +29.9% | DL |
| ±10 bpm Accuracy | 64.7% | 82.8% | +28.0% | DL |
| Euro NCAP 2026 | PASS | EXCELLENT | - | DL |
Deep Learning rPPG-Net model achieves MAE of 6.33 bpm, meeting Euro NCAP EXCELLENT criteria (below 8 bpm).
Experience the real-world performance of our rPPG-Net deep learning algorithm firsthand.
Contact Us