Estimation and Assessment of Upper Limb Movements During Exercises of Children with Musculoskeletal Disorders

摘要

Musculoskeletal disorders can completely take away the possibility of one's locomotion, and in most cases require intensive rehabilitation. Medical services are still one of the least automated, while in the era of increasing emphasis on personalized medicine, the only effective way to overcome most problems can be to automate the rehabilitation process. This paper presents parts of a methodological basis for an automatic expert platform assisting in the process of rehabilitation. We test four machine learning models in tasks that involve assessment of limb exercises and joint rotation estimation, solely based on electromyography signals. In the best case, the models achieved 72% of accuracy in the former, and 0.08 of mean absolute error in the later. The level of errors qualifies these models as acceptable for further development for rehabilitation systems.