Figure 8
From: Arm movement speed assessment via a Kinect camera: A preliminary study in healthy subjects
Classification error rate of the speed types based on a sequential 10% cut of the whole motion. This analysis was carried out to search for the most useful 10% portion of the motion’s signal that distinguishes the speed types. Knowing this percentage can lead to an effective prediction. The two figures represent, from left to right, the intraclass speed classification for f 1 (mean of the instantaneous velocity) and the interclass speed classification for f 2 (standard deviation of the instantaneous velocity) of the hand joint. The intraclass boxplot shows the variation within subjects, while the interclass scatter shows the exact error rate over all subjects. First, we can observe that the classification error diminishes in the middle of the curve; this seems to indicate that the most meaningful section of the motion is at the middle. The smallest error is in the portion 50–60% for both intra- and interclass speed classification.