Real-time detection of unusual behaviour during a physical activity Overview

Regular physical activity (PA) has become one of the most important factors of the health lifestyle. Despite of the problem of initiating PA, i.e. motivating people to increase physical activity after a period of rather sedentary behaviour, there is a big challenge of so called PA maintenance – keeping the people being physically active after the initiating period. One of the most promising methods for PA Maintenance is (remote) monitoring of the physical activity with the goal of observing some physical or physiological parameters, like number of steps, heart rate. In such a monitoring it is very important to detect and understand unusualities in the behaviour, i.e. deviations from the normal/usual behaviour. The unusualities can be defined based on some thresholds (like that normal systolic blood pressure is below 120), but for the most of the vital signals the thresholds (i.e. models) should be personalized, reflecting the personal behaviour (i.e. the past data). This requires learning the models of the usual/unusual behaviour from past data. The main advantage is the possibility to detect some unusualities in the behaviour as soon as they appear, creating an opportunity for a prompt reaction.

This VAS goes one step further by enabling the monitoring in a specific health context, i.e. the monitoring of some parameters that are important for the characterization of a disease (in particular, blood sugar level, weight, BMI, …). A very promising application area for this type of services is the fitness exercising, whereas not only the fitness performances, but also the unusualities in the relevant vital health parameters can be monitored in order to avoid any critical situation.

We argue that this type of services will pave the way for the new generation of the fitness trainings which will maximise the fitness performances but with an active control of the health status of the trainee