Table A1.

Related Research

Study	Recording	Items	Scoring Method	Accuracy
(Eichler et al., 2018)21	- 2x Kinect cameras - $N_S$ = 12, $N_H$ = 10	2	SVM, Single Decision, RF	59%, 91%
(Lee et al., 2018)22	- 1x Kinect camera - Force sensors $N_S$ = 9, $N_H$ = 1	26	Binary rule-based classification	66.7% - 100%
(Otten et al., 2015)23	- 1x Kinect camera - 1x IMU - 1x Glove sensor - $N_H$ = 10	25	SVM and BNN	86% and 93%
(Kim et al., 2016)19	- 1x Kinect camera - $N_S$ = 41	13	PCA and ANN	65% - 87%
(Olesh et al., 2014)20	- 1x LED-marker camera - 1x Kinect camera - $N_S$ = 9	10	PCA	0.03 < $R^2$< 0.98

Abbreviations: $N_S$, stroke subject sample size; $N_H$, healthy subject sample size; IMU, inertial measurement unit; SVM, support vector machine; RF, random forest; BNN, backpropagation neural network; ANN, artificial neural network; $R^2$, correlation coefficient. Accuracy is provided as a percentage compared to manual scores from trained healthcare professionals.