Research On Gait Recognition Method Based On Sensor Information Fusion

Lower limb assisted exoskeleton robot follows human motion and forms a complex coupling system with human body.At present,the research on it mainly focuses on human motion perception and human-machine interaction,motion intention recognition,motion pattern judgment,human-machine coupling control and other aspects.Human gait should be distinguished in order to realize the rapidity,accuracy and stability of exoskeleton system control.The accuracy of recognition directly affects the control effect of the whole system.Therefore,three aspects will be focused: acquisition and analysis of human gait,prediction of human motion signal and recognition of gait phase.(1)Gait information is the basis of research on human motion characteristics,prediction of gait sequence and recognition of gait phase,so a gait acquisition system which can collect gait data completely and accurately is designed.The information of plantar pressure,joint angle and thigh tension is analyzed,and the motion characteristics of different phases in a gait cycle are extracted.(2)Three methods,LMS adaptive filter,weighted zero-order local area method and dynamic neural network are used to predict gait data.For later transplantation to hardware,the performance of three algorithms is evaluated according to the three prediction evaluation index of prediction accuracy,smoothing coefficient and calculation speed.(3)Fuzzy fusion algorithm for gait recognition is used to process multi-sensor information,and sensor information states of different gait phases with inaccurate demarcation and periodicity are described by fuzzy sets.Then the overall gait recognition scheme is designed: firstly,the plantar pressure and the thigh pulling are fused with data;Secondly,the plantar pressure and joint angle are fused with features,and the membership degree of plantar pressure sensor,joint angle sensor and thigh pulling sensor to each phase is obtained by fuzzy reasoning calculation;Thirdly,according to the condition of each sensor information,the hierarchical analysis method is used to carry out weight allocation,and the final recognition result is obtained by global fusion with all sensors.Finally,the confirmatory experiment for gait recognition algorithm is designed,the experimental validation and result analysis of the gait recognition with different walking speed and different experimental objects are carried out,and feasibility and effectiveness of the gait recognition algorithm of multi-sensor fusion based on fuzzy theory is verified.