Abstract Hand and finger injuries account for a significant proportion of cases in emergency departments and require complex, individually tailored rehabilitation. Prescribed occupational therapy to restore hand function is an important therapy method for regaining the ability to work and perform activities of daily living. Since many patients have to perform their exercises at home without direct therapeutic supervision, motivation problems and low therapy compliance are common. In this context, telemedicine-supported occupational therapy offers great potential. As a precursor for a monitoring and feedback tool that increases patient adherence, some of the most important occupational therapy hand exercises are analyzed and classified based on static finger angles recorded using a sensor glove. Recordings from 15 subjects (26.73 ± 6.73 years; 8 men, 7 women; 3 left-handed, 12 right-handed) were divided into training and test data and classified using linear support vector machines. The two trained SVM algorithms can classify the test data with an accuracy of 95.6 % and 93.9 % respectively. For both classifiers misclassifications are relatively rare and only occur mainly between similar hand positions. In addition to static joint angles, other factors such as joint coupling and dynamics should also be taken into account in future work when assessing the quality of performance of occupational therapy exercises.
Waschk et al. (Fri,) studied this question.