Deep learning based blood pressure estimation using ballistocardiogram
Oral presentation
Research approach
Filtered two-channel BCG from chair-mounted PVDF sensors at 4–15 Hz. Combined a 1D CNN and bidirectional GRU to estimate blood pressure without handcrafted pulse features.
View full-size figure
Results
Collected about three hours of resting and post-exercise recordings from each of five participants. Trained and evaluated within each participant using cuff blood pressure as the reference.
All publications