Modeling And Prediction Of Human Gait By Fractional Order Time Series

According to statistics,by 2025,China's population over 60 will exceed 300 million.At that time,China will face a very serious aging problem.With the growth of people's age,the physical function of the people will gradually decrease,and the elderly are prone to fall and hurt because of their lower body function.In addition,the number of lower extremity disability caused by diseases,traffic accidents and industrial accidents is also increasing.Limb injuries greatly reduce the quality of life of these people,and bring some burden to their families and society.In order to improve the physical function of people with lower limb walking disabilities,it is necessary to carry out reasonable rehabilitation training for them.In order to prevent patients from accidentally falling down during training,we need to detect and record the training gait of patients in real time,and obtain the gait information of lower limbs to model and predict.In this thesis,a new gait detection system is proposed,which is composed of rehabilitation training platform and Biometrics series of multi-axis joint angle data acquisition instruments.The gait information of the lower limbs is collected by the angle measuring instrument.Then the data is transmitted to the PC end by the data recorder to display and save.After data acquisition,we need to build a time series model for gait sequence to reflect gait changes.It is found that human gait has obvious long-term memory when walking at slow,normal and fast.However,most of the time series models are integer order short-term memory time series models.In the process of modeling,many effective information are lost due to excessive difference.This kind of model can only reflect short-term memory between sequences,but cannot reflect long-term memory.However,the fractional time series model can avoid the excessive difference to the data,and not only reflect the short-term memory between the sequences,but also reflect the long-term memory.Therefore,this thesis uses the fractional order long-term memory time series model to model the gait data,and combines the Calman filter to predict the gait sequence.It is estimated that we can grasp the abnormal gait of trainees in time and reduce the probability of unexpected occurrence in training.