Locomotion Mode Classification Based on Support Vector Machines and Hip Joint Angles: A Feasibility Study for Applications in Wearable Robotics
Intention decoding of locomotion-related activities covers an essential role in the control architecture of active orthotic devices for gait assistance. This work presents a subject-independent classification method, based on support vector machines, for the identification of locomotion-related activities, i.e. overground walking, ascending and descending stairs. The algorithm uses features extracted only from hip angles measured by joint encoders integrated on a lower-limb active orthosis for gait assistance. Different sets of features are tested in order to identify the configuration with better performance. The highest success rate (i.e. 99% of correct classification) is achieved using the maximum number of features, namely seven features. In future works the algorithm based on the identified set of features will be implemented on the real-time controller of the active pelvis orthosis and tested in activities of daily life.
This work was supported in part by the EU within the CYBERLEGs Plus Plus project (H2020-ICT-2016-1 Grant Agreement #731931) and in part by INAIL within the MOTU project (PPR-AI 1-2).
Andrea Parri, Simona Crea and Nicola Vitiello have commercial interests in IUVO s.r.l., a spin off company of Scuola Superiore SantAnna. Currently, the IP protecting the APO technology has been licensed to IUVO s.r.l. for commercial exploitation.
- 2.World Health Organization: Global Health and Aging (2006). http://www.who.int/ageing/publications/globalhealth.pdf
- 13.Tkach, D.C., Hargrove, L.J.: Neuromechanical sensor fusion yields highest accuracies in predicting ambulation mode transitions for transtibial amputees. In: Proceedings of the 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pp. 3074–3077 (2013)Google Scholar
- 14.Young, A.J., Simon, A., Hargrove, L.J.: An intent recognition strategy for transfemoral amputee ambulation across different locomotion modes. In: Proceedings of the 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pp. 1587–1590 (2013)Google Scholar
- 16.Pratt, G.A., Williamson, M.M.: Series elastic actuators. In: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 399–406 (1995)Google Scholar
- 19.Chen, P., Liu, S.: An improved DAG-SVM for multi-class classification. In: Proceedings of the 5th International Conference on Natural computation, pp. 460–462 (2009)Google Scholar
- 21.Parri, A., Yuan, K., Marconi, D., Yan, T., Munih, M., Molino Lova, R., Vitiello, N., Wang, Q.: Real-time hybrid ecological intention decoding for lower-limb wearable robots. IEEE Trans. Mechatron. (Accepted for publication)Google Scholar
- 22.Jang, J., Kim, K., Lee, J., Lim, B., Shim, Y.: Online gait task recognition algorithm for hip exoskeleton. In: Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 5327–5332 (2015)Google Scholar