A smartphone based real-time daily activity monitoring system

A real-time activity monitoring system within anAndroid based smartphone is proposed and evaluated. Motionand motionless postures may be classified using principlesof kinematical theory, which underpins hierarchicalrule-based algorithms, based on accelerometer and orientationdata. Falls detection was implemented by analyzingwhether the postures classified as ‘lying’ or ‘sit-tilted’ postureare deemed normal or abnormal, based on the analysisof time, users’ current position and posture transition. Experimentalresults demonstrate that the approach can detectvarious types of falls efficiently (i.e., in real-time within asmart phone processor) and also correctly (95 % and 93 %true positives for falls ending with ‘lying’ and ‘sit-tilted’ respectively).The approach is reliable for different subjectsand different situations, since it is not only based on empiricalthresholds and subject-based training models, but inaddition it is underpinned by theory.