Wearable full-body inertial measurement with task classification using Deep Learning


  • Jeremías Gaia Amorós Instituto de Automática UNSJ-CONICET
  • Eugenio C. Orosco
  • Carlos Soria


Embedded systems, Inertial Measurement, Non-invasive sensing, Deep Learning


In this work, an embedded system is developed for the non-invasive sensing and storage of biomechanical variables of people. It takes advantage of wearable technology, distributing sensors in strategic points of the body, ergonomically and functionally. The results are verified by recording and analyzing tasks performed by six subjects to form a database. These tasks include being stood up, sitting down or standing up from a chair, going upstairs and downstairs and walking. Additionally, a convolutional neural network is tested for offline task classification. This work aims to initiate a process that ends in assistance-oriented applications, for the development of better injury rehabilitation techniques and support for elder people, among others. In this way, it seeks to open a path towards an improvement in the living conditions of people with and without reduced activities of daily living capacity.


Download data is not yet available.



How to Cite

Gaia Amorós, J., Orosco, E. C., & Soria, C. (2021). Wearable full-body inertial measurement with task classification using Deep Learning. IEEE Latin America Transactions, 19(1), 115–123. Retrieved from https://latamt.ieeer9.org/index.php/transactions/article/view/2718