“…Similar to our previous study [ 13 ], the system consisted of a sensor, user interface, and scoring algorithm. A depth sensor, Kinect V2 (Microsoft, Redmond, WA, USA) was used to extract motion features ( Figure 2 ), which is an inexpensive and easy to use sensor that has been widely applied in the rehabilitation area [ 13 , 15 , 16 , 17 , 32 ], and whose effectiveness was verified in our previous FMA studies [ 13 , 17 ]. The system has the same user interface as our previous automated FMA system [ 13 ], which provided an instruction video that was prerecorded by a well-experienced clinician.…”