%L repoupi144322
%O SINTA ID DOSEN PEMBIMBING
Agus Heri Setya Budi: 6003446
%I Universitas Pendidikan Indonesia
%X Falls among the elderly are a serious health issue, where current solutions
are often hindered by the weaknesses of cloud-based detection systems, such as high
latency, significant power consumption, and internet dependency. This research
aims to optimize a fall detection system by developing and testing a hybrid
architecture based on edge computing on an Artificial Intelligence of Things
platform using the ESP32-S3 microcontroller. The research method is Research and
Development (R&D), wherein this study builds a functional wearable prototype
using an MPU6050 sensor and an ESP32-S3 microcontroller. It implements a
hybrid detection algorithm, which combines a threshold-based method as an initial
trigger and a Random Forest classifier for final validation, to run entirely locally
on the device. The study trains the Random Forest model offline using the public
SisFall dataset and tests it through a series of simulated fall scenarios and Activities
of Daily Living (ADLs) in a controlled laboratory environment. The results show
that this edge-based system achieves a fall detection success rate of 91.25% with a
very low false positive rate of 2.5%. The end-to-end system response time, from the
moment of the incident until the notification is received, averages 4.85 seconds,
which demonstrates the superiority of the edge computing architecture in terms of
speed and efficiency compared to cloud-based approaches. This system offers a
more responsive, reliable, and power-efficient solution.
Keywords: Fall Detection, Edge Computing, Artificial Intelligence of Things,
ESP32 microcontroller, Random Forest.
%D 2025
%A - Galuh Yudha Prastyo
%A - Agus Heri Setya Budi
%K Deteksi Jatuh, Edge Computing, Artificial Intelligence of Things, 
Mikrokontroler ESP32, Random Forest
%T OPTIMASI SISTEM DETEKSI JATUH UNTUK LANJUT USIA
BERBASIS EDGE COMPUTING MENGGUNAKAN ALGORITMA
HYBRID PADA PLATFORM ESP32