Implementation of a real-time human movements classifier by using mobile equipment

This paper deals with implementation of a real-time system for classification of human movements based on the use of information recorded through an Android smartphone, namely acceleration and heart rate signals. One of the main features of the application is that the overall equipment is cost effective, since triaxial accelerometers are currently widely available on Smartphone without additional costs. Similar applications previously appeared in literature, considered ad-hoc devices both to measure and communicate data. The classification is carried out based on the use of 5 classification indexes, appropriately computed from recorded signals. The present version of the Android application allows discriminating among 9 classes of movement and can classify individual movements or operate in continuous mode. Results shows that the classification based on strictly acceleration signals is extremely effective, while it is slightly less accurate discrimination between pairs of movements whose discrimination is related exclusively to heart rate. The continuous operation mode could be interesting for the remote surveillance of individuals (e.g. elderly and/or disable people).