Identification of Human Walking Patterns Using 3-D Dynamic Modeling

One of the most common activities of our day to day life is walking. However simulating a human walking motion is one of the most difficult tasks to accomplish. Inherently it is an inverted pendulum like system and involves a large number of degrees of freedom. In this thesis we have modeled the human walking motion. The system is designed using a human body model in the form of a kinematic chain consisting of rigid links and revolute joints. Human walking patterns contain information like identity, presence of physical disability and loading conditions of a person like carrying a backpack. We have extracted some of these information and have used our model to discriminate various walking motions. The information that we have used are joint torque and angle sequences modeled using ARMA modeling and Dynamic Time Warping. Our human walking model is validated by comparing it with Stanford marker data