The Gait Analysis And Application In NDDS Classification Based On Deterministic Learning

Gait is the human body posture when walking, which is a kind of complex movement process. Human gait need the coordination by the multiple systems like the nervous system, the sensory system and motor system at the same time, so in some ways the body posture can reflecte the body's biological characteristics, health, etc. Gait analysis is a study method of walking, which is an effective means of evaluating walking system by objectivly recording of walking and revealing the key link in the process of human gait and the influence factors through biological mechanics and kinematics method. Because gait feature can be accessed in the distance, so from the point of view of intelligent video surveillance, gait recognition has a very wide range of application prospects. Due to the different disease can have different special abnormal gait, such as neurological degenerative diseases, including Parkinson's disease(PD), Huntington's disease( HD), and Amyotrophic Lateral Sclerosis(ALS). So the quantitative analysis of gait also may provide important clues for the diagnosis of these disease.The deterministic learning(DL) theory proposed in recent years is a new machine learning theory, which provides systematic design approaches for knowledge acquisition, representation, and utilization in uncertain dynamical environments in unknown dynamic environment. Locally-accurate identification of the gait system dynamics is achieved by using radial basis function(RBF) neural networks(NNs) through deterministic learning. The obtained knowledge of the approximated gait system dynamic is stored in constant RBF networks. A bank of estimators are constructed using constant RBF networks to represent the training gait patterns. In the testing phase, by comparing the test gait pattern with this group of dynamic estimators, a set of recognition errors are generated, and the average L1 norms of the errors are taken as the similarity measure between the dynamics of the training gait patterns and the dynamics of the test gait pattern. Therefore we can quickly recognize the test gait pattern according to the smallest error principle.Based on the DL theory, and the previous research achievements of our team, this dissertation has carried on the exploration and research of gait analysis in practical application. Based on the existing related theory and algorithm of the gait feature extraction, we try to use a new method of gait feature extraction.Then we mainly introduced the design details of gait quick- recognition system.The system involves the data acquisition, gait training, and gait recognition framework design, and intuitivly demonstrates the process of the gait recognition and the recognition results. The development of the system for gait recognition and neurodegenerative disease classification is carried out on Matlab GUI. The system is aim to improve the recognition rate and recognition speed of human gait, and the success rate of disease classification. In order to improve the computing speed of neural network, parallel computing tool(Jacket),which adopted nvidia parallel computing architecture(CUDA) is used to accelerate parallel computing process.