The Application Of Nonlinear Identification In Human Dynamic Balance Modeling

Balance which is the most important function of human body plays an important role in human life, when who maintains a variety of static posture, completes a variety of simple or complex movements, makes various sports activities. As for the elderly, patients with cerebral apoplexy and disability patient, balance becomes more important. Therefore, the study of human balance is of great significance.The modeling of human dynamic balance, can be used to analyze the ability of restore balance in the external disturbance. Existing studies usually used PD / PID controller as the central nervous system, to control the linear model of human body, such as, an inverted pendulum model or a double inverted pendulum model, under the small angle shaking, and at last, get the body homeostasis curve adjustment process. However, that curve is not a true reflection of the human balance adjustment process, because of the nonlinear system of human body.The nonlinear system identification methods can solve the problem of nolinear modeling well. In recent years, many new identification methods base on neural networks, fuzzy logical algorithm and genetic algorithms have become a current research trend in nonlinear system identification area. Thus, the main research contents of this paper include the following aspects:Firstly, this paper based on six degrees of freedom motion platform, studies the frequency characteristics of the process of homeostasis in the body to get human bandwidth under passive motion mode. In this paper, an acceleration sensor is used to obtain the body center of gravity acceleration data, solving the problem of people do not accurately reflect the mass trajectory center of body when body center of pressure changes with frequency. Well reflect the body’s balance of properties and get the body bandwidth of homeostasis adjustment process.Secondly, this paper uses the nonlinear system identification method, which is based on genetic algorithm for optimizing the BP neural network, to model the dynamic balance of human body. At the same time, the weights and thresholds of different individuals in the process of balancing adjustment are obtained, which lays the foundation for the classification of human body balance ability.Thirdly, this paper puts forward a new classification method of balance ability, according to the nonlinear system identification results. The specific method is classification according to weighted values and thresholds of different individuals by using support vector machine(SVM), classification standards is based on the time of human body in dynamic balance adjusting process. Finally this paper realized the classification for human balance ability.In conclusion, this study adopts the passive control mode to get the frequency characteristics during the processing of body dynamic balance, the results are more objective and accurate. Based on nonlinear system identification method, this paper has obtained the body dynamic balance process model. Using support vector machine(SVM) realizes the classification for human balance ability with a high accuracy, reached 94%. This research is not only provide data support for medical research but also provides a new approach for the classification of human balance ability.