Multi-task non-contact ballistocardiogram-based vital signs monitoring in acupuncture

This paper introduces an innovative non-contact vital signs monitoring system for acupuncture patients, combining multi-channel ballistocardiogram (BCG) signals with multi-task deep learning for simultaneous heart rate estimation, respiration rate estimation, and posture detection.

Paper

Abstract

The study introduces an innovative approach for efficient vital signs monitoring in acupuncture by combining multi-channel ballistocardiogram (BCG) signals and multi-task learning, leveraging the polyvinylidene fluoride (PVDF) film sensor and deep neural networks. The proposed system utilizes non-contact under-mattress BCG signals for heart rate (HR), respiration rate (RR) estimation and lying posture detection. Integrating Gated Recurrent Unit (GRU) and Multi-head Self-Attention (MHSA) mechanisms, the system achieves 98.7% accuracy for activity recognition and 97.6% for lying posture classification. The best MAE for HR achieves 0.77 bpm and for RR achieves 0.43 brpm, compared to an FDA-approved device.

Key Contributions

• Wireless and non-contact BCG-based system for vital sign monitoring
• Hybrid deep learning model (GRU + MHSA) enabling multi-task detection
• High accuracy across different testing scenarios with 25 participants
• Safety monitoring solution for acupuncture patients without interfering with acupuncture points

Recommended citation: T. T. Vo, Q. P. Le, T. N. Nguyen, J. Choi, S. Mondal, B. Lee, and J. Oh, "Multi-task non-contact ballistocardiogram-based vital signs monitoring in acupuncture," *Computers in Biology and Medicine*, vol. 203, p. 111461, February 2026, doi: 10.1016/j.compbiomed.2026.111461.