Transportable and scalable system for activities and exercises recognition in real-time
Kévin Chapron, Kévin Bouchard, Élise Duchesne, Sébastien Gaboury · 2017
Patients concerned with neuromuscular disorders are usually given a program of multiple physical exercises to do. These exercises must be recognized by devices to ensure the therapist that the patient did it well. To this end, we developed a specific method allowing the recognition of two types of activities on different window sizes. The method consists of two devices communicating over Wi-Fi, to use both wearable components and powerful embedded computer. Data is collected during fixed windows by the first device. Then features are calculated on the retrieved data for the recognition algorithm. Participants took part in an experiment based on four physical exercises and six daily activities. The method induced good results.