A neural network based application for remote monitoring of human behaviour

The daily behavior of an individual is an important indicator to identify and prevent health problems, so that some works are focusing their investigations on techniques to carry out remote and continuous monitoring of such human behaviors. In this context, our work presents a neural network based application for the automated classification of human movements. This application identifies six basic movements, from the signals obtained from a mobile triaxial accelerometer, and uses these movements to figure out behavioral patterns from their uses. The set of features used to characterize the accelerometer curves provided a classification accuracy of about 91%. Another contribution of this study was to raise up important practical issues that must be considered as extensions of this kind of intelligent monitoring application.