Formation of a Dataset for Assessing Fine Motor Skills of Handwriting after Stroke and an Example of Its Application

As part of the work, a study of the handwriting of patients after a stroke during the rehabilitation process is being carried out. The use of simple figures (straight and wavy lines, a spiral, a triangle, the short phrase “She was given tea”) is analyzed. The formation of multiple sessions of handwriting collection, its parametrization and the resulting structure of the data set are described. The set was formed by taking handwriting samples from healthy people and patients after a stroke. An example of solving the problem of classifying handwriting into two classes is given: before and after a stroke. To classify the parameters into two classes, ensemble machine learning methods and main types of classifiers were selected, such as k-nearest neighbors (KNN), support vector machine (SVC), decision tree (DT) and naive Bayes classifier (NB). This approach can be applied in the future when assessing fine motor skills during the rehabilitation period based on the metric of proximity to the class before the stroke.