Internet des Objets


ACTIvity as an antiDOTE

Collaborators: David Da Silva Andrade, Dr. Laurent Gantel,  Prof. Andres Upegui

Physical inactivity has been identi fied as a major contributor to the exacerbation of physical illnesses. The WHO identi fied it as the fourth leading risk factor of global mortality after high blood pressure, tobacco use and high blood glucose. Therefore, in recent years, many actions against inactivity have come to the fore. For instance, diverse pedometer devices have been developed to help people reach certain physical activity goals, like walking 30 minutes per day. However, an equivalent recommendation for disabled people using wheelchairs is missing and the few studies that have dealt with this issue concluded that commercial physical activity measurement devices are not appropriate for them. This project has the objective of developing an embedded physical activity measurement system for disabled people using wheelchairs, by exploiting on-body and wheelchair-mounted wireless sensors. This project gathered together data scientists (Pr. Perez-Uribe), embedded systems designers (Pr. Upegui & Pr. Giandomenico), biomechanics experts (Pr. Schmitt) and Human motricity and handicap experts (Pr. Degache).


During the first phase of the project, we used o ff-the-shelf sensors to capture data and apply feature-extraction and machine learning techniques to the sensor readings in order to come-up with activity classes, associated with activity intensities (e.g., light, moderate and vigorous). In parallel, we developed our own embedded hardware to optimize size, maximize comfort, and minimize costs. Diverse activities like resting, deskwork, and wheelchair propulsion along di fferent surfaces and slopes were considered.


During the second phase, with the aim of developing regression models of energy expenditure, (in collaboration with the Institut des Sciences du Sport de l'Universite de Lausanne) we recorded users on a treadmill performing wheelchair propulsion on ramps of di fferent slopes and speeds. The results of this project are a) an embedded system for disabled self-monitoring of the physical activity, and b) a series of machine learning models, one based on activity recognition and the second one on regression equations, to estimate the users energy expenditure.


Even if the system still needs further tests, in particular with disabled people, this prototype and its future development aims at closing a gap regarding the availability of self-tracking/motivational devices among disabled people.

© 2018 | hepia - LSN Laboratoire de Systèmes Numériques, rue de la Prairie 4, CH-1202 GENEVE