Experimental Study: Deep Learning-Based Fall Monitoring among Older Adults with Skin-Wearable Electronics

Area of Science:

• Gerontology
• Biomedical Engineering
• Computer Science

Background:

• Older adults face increased fall risks due to aging, leading to significant healthcare burdens.
• Current automatic fall detection systems for the elderly are insufficient.
• Falls in older adults represent a major medical and societal challenge.

Purpose of the Study:

• To develop and validate a wireless, skin-wearable electronic device for motion sensing.
• To create a deep learning algorithm for reliable fall detection in older adults.
• To assess the efficacy of different device placements and datasets for fall detection.

Main Methods:

• Designed and fabricated a cost-effective, flexible, skin-wearable device using thin copper films with a six-axis motion sensor.
• Collected motion data from older adults performing various activities.
• Investigated deep learning models, device placement (e.g., chest), and dataset variations for classification.

Main Results:

• Achieved over 98% accuracy in fall detection when the device was placed on the chest.
• Demonstrated the feasibility of using a skin-wearable device for accurate motion data collection.
• Highlighted the necessity of large, elder-specific motion datasets for optimal performance.

Conclusions:

• A wireless, skin-wearable motion monitoring device combined with deep learning offers a promising solution for fall detection in older adults.
• Optimal device placement (chest) and comprehensive datasets are critical for high accuracy.
• Further development using elder-specific data can enhance the reliability and applicability of this technology.