INFOrM - Uma abordagem para detecção de quedas baseada em sensores de movimento infravermelhos e acelerômetros

Abstract

The increase of life expectancy and higher quality of life has caused the aging of the world population. A fairly common problem for the elderly are falls, which can be fatal in some cases. In such scenarios, technology should offer support for better monitoring of this age group. With the Internet of Things (IoT), anything can be accessed anytime, anywhere. Through the information of these things an Ambient Assisted Living (AAL) System can help people maintain their independence and live better. One of the functionalities of this type of system is to detect falls. This master dissertation paper presents INFOrM (INdoor Fall detectiOn Method), a method for fall detection that uses machine learning to generate a classification model based on high and low level information, from accelerometers and passive infrared sensors. The classifier uses technology that is easy to accept by the user and has redundant and complementary information that can improve system detection when analyzed together, but can also be used independently, avoiding system shutdown. The results obtained from experiments show that the combined use of sensors that respect the privacy of their users, such as motion sensors and accelerometer, can achieve high accuracy rates in fall detection, including slow falls and risk scenarios, such as the bathroom.