Energy-Aware IoT-Based Method for a Hybrid On-Wrist Fall Detection System Using a Supervised Dictionary Learning...

Area of Science:

• Engineering
• Computer Science
• Gerontology

Background:

• Falls are a significant health concern for the aging population.
• Existing Internet of Things (IoT)-based fall detection systems face energy consumption challenges.
• Wearable, wrist-based solutions offer comfort but require efficient power management.

Purpose of the Study:

• To propose a novel, energy-aware IoT architecture for wearable fall detection.
• To develop a gateway-less monitoring system using Message Queuing Telemetry Transport (MQTT).
• To enhance fall detection accuracy and reduce energy consumption for long-term use.

Main Methods:

• Implemented a hybrid double prediction technique based on Supervised Dictionary Learning.
• Utilized a controlled dataset for offline training and real-world measurements for online validation.
• Developed an energy-aware IoT architecture for MQTT-based, gateway-less monitoring.

Main Results:

• Achieved high offline (99.8%) and online (91%) fall detection accuracy.
• Demonstrated significant battery consumption optimization, extending working hours by a minimum of 27.32 hours.
• Outperformed most related works relying solely on accelerometer data.

Conclusions:

• The proposed system offers a promising solution for reliable, long-term, anywhere-anytime fall detection.
• The energy-aware architecture addresses critical power consumption concerns in wearable IoT devices.
• This approach enhances the feasibility of wearable fall detection for practical applications, especially for the elderly.