FULL-BODY KINEMATICS
Dr . Asbeck and his lab have created and released a new dataset of natural motion , including kinematics of workers in a store and individuals doing unscripted activities of daily motion . This is the largest dataset of real human motion to date , and is the only one that captures how people move in everyday life . Using this dataset , Dr . Asbeck and graduate student Jack Geissinger have used machine learning algorithms to predict the full-body kinematics using only five or six inertial measurement units : one on the pelvis and / or chest , and one on each of the wrists and ankles . Even with this minimal sensor suite , the algorithms have surprisingly accurate results . Dr . Asbeck and his team are now using similar algorithms for monitoring progress during stroke rehabilitation .
Alan Asbeck Assistant Professor
The left panel shows several postures predicted by our algorithms (“ Predicted ”) as compared to the true posture (“ Ground Truth ”). The right panel shows the locations of the sensors ( white and blue circles ) used to perform the motion inference .
28 RESEARCH • VIRGINIA TECH MECHANICAL ENGINEERING ANNUAL REPORT • 2019-2020