Nexar: sistema de detecção de queda de baixo custo para idosos

Abstract

The high incidence of physical falls among older adults and the limitations of often expensive commercial devices motivate the development of more accessible and robust wearable solutions. This work proposes an automatic physical fall detection system based on a low-cost device using ESP32 and six-axis inertial sensors, capable of collecting raw acceleration and angular velocity data. The solution adopts a three-layer architecture: embedded preprocessing on the device, responsible for filtering, normalization, and preliminary identification of suspicious events; a REST API developed in Spring Boot, which centralizes records and forwards the processed data; and a classification module in Python, in which deep learning models based on CNN and LSTM distinguish real physical falls from daily activities. The system is also integrated with a mobile application that enables real-time alerts and uses the smartphone’s GPS to record the location during the detected event. The approach follows incremental prototyping and experimental testing with real data. The expected results include greater detection accuracy, reduced false alarms, and improved efficiency in continuous monitoring, making the system a viable and affordable alternative to existing solutions.