Abnormal gait detection and classification using depth camera

Abstract

This research aims at developing a method to detect abnormal gait from depth images and to classify abnormal gaits of patients. Recently, motion capture system is popular used in the analysis of human gaits. However, a motion capture system remains many weaknesses such as costly and complicated set up, and requiring professional technicians to manage the motion capture system. This work introduces a new approach to detect and classify abnormal gaits by using depth images and skeleton joints of the human subjects detected from the images. The system feeds the data including depth images and positions in 3D of skeleton joints into a hidden Markov model as well as K-means clustering to approach a new effective solution to replace conventional motion capture system. We tested our approach with a large number of subjects to validate its performance and shown that the proposed our system performs well. Therefore, this system may be applicable to help doctors in medical diagnosis and treatment process.