Gait pattern recognition and control using a neural network model and PSO method

In this thesis, we investigate the development of body models using artificial neural networks (ANN) with the gait data measured by the VICON motion capturing system. The models will then be used to develop controllers that maintain the body's center-of-mass (COM) during gait. More specifically, we first use leg and arm motion data to build the body model using an artificial neural network (ANN), which simulate a human's balance dynamics. Second, we develop an inverse control using the gait data that represents the person with an injured leg and feed into the model that generated the COM to analyze the COM of the person with an injured leg. Third, the Particle Swarm Optimization Method (PSO) is used to design a controller which finds the optimal motion for the affected or injured right leg in order to maintain body balance. The PSO was used to optimize the center-of-mass in this research based on the right leg position values which help maintain full body balance. Lastly, we show that the balance model we established can be used for gait pattern recognition and identification, which can help distinguish gait among individuals, thus forming a type of identification.;The significance of this research is geared toward physical rehabilitation; establishing a database of gait pattern for populations with medical ailments such as diabetes which will help enable precise diagnosis to help correct body motion. This research also helps in the development of a walking-aid device for those subject to lower extremity injuries.