Ambulation Assessment Using Depth Cameras

摘要

This work outlines a method for patient ambulation assessment based on tracking the human body using a depth camera. To classify static postures and dynamic movements, a hierarchical classifier is proposed. By analyzing the relative positions of the tracked joints, key features are extracted and input into seven Support Vector Machine classifiers. The performance of the SVMs show that the average accuracy of the entire hierarchical classifier is 96%. For the dynamic movements, an average accuracy of at least 90.27% was observed for distance traveled.