A neural network based fall detector

• Autor(es): Rodrigues, Pedro João ; Amaral, J.S. ; Igrejas, Getúlio
• Data: 2010
• Identificador Persistente: http://hdl.handle.net/10198/4829
• Origem: Biblioteca Digital do IPB
• Assunto(s): Fall detector; Neural network
• Tipo de Documento: Objeto de conferência
• Idioma: Inglês
• Contribuidor(es): Biblioteca Digital do IPB

Descrição

In this project we present an intelligent fall detector system based on a 3-axis accelerometer and a neural network model that allows recognizing several possible motion situations and performing an emergency call only when a fall situation occurs, with low false negatives rate and low false positives rate. The system is based on a two module platform. The first one is a Mobile Station (MS) and should be carried always by the person. An accelerometer is implemented in this module and its information is transferred via a radio-frequency channel (RF) to the Base Station (BS). The BS is fixed and is connected to a GSM (Global System for Mobile communication) module. A neural network model was built into the BS and is able to identify falls from other possible motion situations, based on the received information. According to the neural network response the system sends a SMS (Short Message Service) to a destination number requesting for assistance.