Systematic Research On Human Gait Phase Recognition Based On Multi-sensor Information

In recent years,wearable exoskeleton has become a hot research topic at home and abroad.It can assist the human body to walk,enhance the human body's movement ability and expand the body's load capacity.In order to control the execution process of a wearable exoskeleton control system quickly,effectively and stably,it is necessary to study the gait phase recognition of the wearer and the exoskeleton.The accuracy of the recognition result will directly affect the performance of the whole exoskeleton control system.Therefore,it is very important to study a gait phase recognition system with high accuracy to control the exoskeleton robot.In this paper,based on the multi-sensors,the foot pressure and the position information of the joints of the lower limbs are collected during the walking of the lower limbs,and then the gait of the human body is identified,It involves the analysis of human lower limb motion mechanism,the research of the human gait phase recognition algorithm,the design of gait phase recognition scheme for human lower limb,and the analysis and the acquisition,analysis and processing of lower limb motion data.The main research contents of this paper are as follows:Firstly,the movement mechanism of human lower limb walking is analyzed,and then the gait phase division based on the rationality of the human gait phase division in the process of the execution of the exoskeleton control system is realized,so that the exoskeleton can be controlled and executed better.Then,the algorithms related to human gait phase recognition are studied and analyzed,including fuzzy logic control algorithm,support vector machine,and a genetic algorithm is selected to optimize the parameters of recognition algorithm,as well as a brief overview of data fusion technology.Secondly,three human gait recognition schemes are designed:gait phase recognition based on plantar pressure,attitude information and identification based on foot pressure fusion attitude information.For the first scheme,the pressure distribution of the lower extremities during walking is analyzed,and the data on the sensitive area of the sole force are collected by the foot pressure acquisition system of the lower limbs during the walking process.According to the characteristics of foot pressure data,fuzzy logic,inference algorithm which is optimized by genetic algorithm are used.For the second scheme,the motion data of the lower limbs of the human body are collected using the attitude data acquisition system,According to the characteristics of the information curve of the lower limb posture,a moving window detection algorithm is proposed to segment and identify the gait phase in the walking process of the human's lower limb joints.For the third scheme,the data acquisition system is designed by using a wearable shoe,inertia attitude module,a control chip,and the data information of human lower limb walking is collected synchronously.By fusing plantar pressure and ankle attitude information,support vector machine(SVM)and genetic algorithm(GA)optimized support vector machine(SVM)are used to classify and recognize human gait phase.Finally,according to the analysis of the three gait phase recognition schemes designed in this paper,it is concluded that all three schemes can divide and recognize the gait phase of the human body.However,in the third scheme,both the analysis and processing of the data and the accuracy of the last gait phase identification are better than the first two schemes in recognition effect and feasibility.