Elder Falls Detection Based on Artificial Neural Networks

Abstract

A study of Artificial Neural Networks (ANNs) in the elder falls detection problem is proposed. There are many efforts trying to provide an independent life for the elderly people. Fall event is one of the main problems that affect people in this age group. In order to provide a comfortable solution of this problem for elderly people, this paper presents an implementation of falls detection in mobile phones based on ANNs, because many smartphones have accelerometers inside them and they are not so inconvenient for elder to carry it. Besides, this work contains a performance comparison among three types of ANN (Multilayers Perceptron, Radial Basis Function and Kohonen ANN). It is also shown the process of falls database creation used in this study, obtained from acceleration signals of a mobile phone running Android operating system. The experimental results show the efficiency of each ANN by specificity and accuracy parameters.