Mejora de solución tecnológica con TinyML para detectar y notificar caídas y situaciones de riesgo

Abstract

Response time in emergency situations such as falls, car accidents, and insecurity problems is critical; lack of immediate attention can worsen injuries, cause fractures, contusions, irreversible damage, and prolonged recovery periods. This project focuses on developing an improved technological solution for the detection and notification of falls and risk situations using TinyML. The project aims to develop new functionalities in the mobile application and execute a test plan to validate the system's operation. The design of the wearable device was improved, including a longer-lasting battery. Furthermore, a multi-platform architecture was implemented, integrating a Flask backend with MySQL, a React web dashboard, and a Flutter mobile application. The system incorporates multi-channel notifications (SMS, WhatsApp, email, push notifications) and a dynamic response time mechanism to discard false alarms. The test plan results showed high satisfaction: 4.8/5 and 4.6/5, confirming the correct generation of alerts. The project establishes a competitive alternative to commercial solutions. Keywords: Wearable device, edge computing, emergency alert, real-time monitoring