Two mandatory conditions in the development of tele-rehabilitation platforms are: (i) being based on affordable technologies and (ii) ensuring the patient is performing the exercises correctly. To do so, the present study proposes a cognitive algorithm based on a Hidden Markov Model (HMM) approach to assess in real-time the quality of a human movement recorded through a low-cost motion capture device. The assessment of the correctness of the exercises, which includes the detection of multiple undesirable compensatory movements, shows a very high accuracy (the average performance = 97%). In addition, the proposed model shows a potential for providing the patients with real-time feedback on their performance (up to five times a second). A certain limitation of the model occurs for the compensatory movements characterized by an absence of translational motion of the centre of mass (17% of misclassifications). In this situation, additional features are required to properly assess the quality of the therapeutic exercise.
|Number of pages||19|
|Journal||International Journal of Artificial Intelligence|
|Publication status||Published - 1 Mar 2018|
- Hidden Markov models
- Real-time motion assessment
- Rehabilitation exercises