Identification of People in a Household Using Ballistocardiography Signals Through Deep Learning.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Wearable Technology

Background:

• Traditional health monitoring sensors often require skin attachment, limiting comfort and daily usability.
• Developing non-adhesive, non-invasive sensors is crucial for unobtrusive health monitoring in older adults.
• Polyvinylidene difluoride (PVDF) piezoelectric sensors offer a promising avenue for capturing vital physiological signals without skin contact.

Purpose of the Study:

• To develop a novel health monitoring system utilizing a non-adhesive PVDF piezoelectric sensor.
• To explore the potential of ballistocardiography (BCG) signals for individual identification.
• To enable continuous health monitoring while allowing individuals to maintain normal daily activities.

Main Methods:

• Acquisition of vibration signals (BCG) from ten subjects using a PVDF piezoelectric sensor.
• Construction and training of a neural network model using 252 acquired signal cases.
• Testing the neural network for the identification of five individuals based on household assumptions.

Main Results:

• The developed system achieved good classification probabilities and accuracy rates across all 252 tested cases.
• High classification accuracy was observed in nearly all instances of individual identification.
• The system demonstrated effective identification capabilities for a group of five individuals.

Conclusions:

• The PVDF piezoelectric sensor-based system shows significant promise for accurate individual identification.
• The system's ability to identify individuals based on BCG signal frequency components is validated.
• Future work should address daily signal variations to further enhance the system's robustness for long-term monitoring.