Research On Algorithm Of Lower Limb Movement Pattern Recognition And Motion Prediction

The purpose of the study on intelligent lower limb prosthesis is improving and raising the lower limb disabled people’s life quality, and promoting the development of China’s medical and social harmonious and stable. At present, there have been appeared lower limb prosthesis which have highly intelligent bionic degree at home and abroad, but the price is still higher, it is difficult to spread to the majority of people with disabilities. Therefore, increase the intensity of research and development of high performance, low cost intelligent artificial leg product have great significance to improve the daily lives of people with disabilities in our country.The lower limb movement’s pattern recognition and the accurate prediction of the lower limb motion are the premise research of intelligent lower limb prosthesis. This paper is mainly based on lower limb motion information acquisition, exploration and research theextension of Neural Network algorithm in the field of intelligent artificial leg. To achieve the human lower limb motion’s pattern recognition and the accurate prediction of the lower limb motion based on theNeural Network algorithm,the research mainly do the following points:(1) In order to better achieve the human lower limb motion pattern recognition and motion prediction,we analyzed the status and characteristics of human lower limb motion parameters in detail.And then set up the knee angle’s acquisition system of human lower limb motion.Use the angle’s mean ratio method to achieve simple normalized of the angle of knee joint signals,and change the knee joint angle signal into the characteristics of knee joint angle.(2) In the motion state’s pattern recognition, this paper chooses the BP Neural Network algorithm and two improved algorithms of BP and also self-organized competitive Neural Network. We use these four kinds of network established pattern recognition model of multi state of motion,and then train and simulate these models. Finally through the recognition result,we find that the self-organized competitive Neural Network model have a better identification accuracy.It’s faster and more stable than others.(3) We introduce another kind of Neural Network algorithm of RBF Neural Network, and then use L_M back propagation algorithm of BP Neural Network and RBF Neural Network to establish the lower limb posture’s prediction model, forecast for human lower limb motion. Comparing the simulation results of two kinds of model, we find that the human lower limb motion prediction model based on RBF Neural Network have a precise prediction, almost coincide with the actual movement trend,and applicable to the human lower limb motion prediction.