Fig 1. We used a three-step workflow to measure movement frequency from smartphone videos.

Participants first provided a video recording of themselves performing the task of interest (we assessed five different tasks in this study: finger tapping, hand opening/closing, hand pronation/supination, toe tapping, and leg agility or heel tapping). Next, we used OpenPose to estimate body and hand keypoint locations on each video. We then developed post-processing software to calculate movement frequency using the OpenPose outputs.