Wearable IoT Devices: Feasibility and Effectiveness in Implementing Local Decision-Making for Elderly Fall Detection

Context: The growing use of IoT wearables in healthcare highlights their potential for elderly fall detection. However, wearable devices based on inertial sensors have processing and energy consumption limitations, making them unsuitable for complex algorithms. Traditional cloud systems face latency and bandwidth issues when processing inertial data, which may delay caregiver responses. Edge computing, with local decision-making, emerges as an efficient and energy-conscious alternative. Problem: The overload in transmitting inertial sensor data to remote servers can delay notifications, putting elderly individuals whose falls are undetected at risk. Rapid intervention is crucial to prevent severe consequences. Solution: This study implements a decision tree algorithm processed locally on the ESP32 microcontroller, enabling real-time fall detection while reducing energy consumption and response times. SI Theory: The study applies General Systems Theory by integrating inertial sensor data, software algorithms, and human behavior. This systemic approach optimizes device performance, promoting greater efficiency in real-time health monitoring and improving elderly safety. Method: This applied, quantitative research pre-processed accelerometer and gyroscope data into statistical metrics. A decision tree algorithm was developed and validated for local processing on the ESP32, using a dataset tailored for fall detection. Summary of Results: The proposed solution on the ESP32 detected falls with precision, without false positives, distinguishing daily activities in controlled tests. Additionally, it demonstrated high energy efficiency, validating its viability for IoT wearable devices. Contributions and Impact in the IS area: This study validates the integration of edge computing with wearable devices, balancing computational efficiency, energy consumption, and safety, offering effective solutions for health monitoring.