Research On Human Lower Limb Motion Based On Multi-sensor Information

One of the key issues facing the research of intelligent power lower limb prosthesis is to realize the recognition of the wearer's motion intention by the prosthesis,and provide the user with corresponding control strategies and action assistance.The information commonly used to predict the intention of human movement mainly includes: neural signals,mechanical signals,and biomechanical signals.Considering that a single signal is not enough to provide all the information for judging the gait pattern,this article is based on the human lower limb movement data sensing collection system,using sensors with different physical information for gait pattern recognition.The human body lower limb movement data sensing collection system proposed in this paper includes a plantar pressure collection system and a large and large leg movement information collection system.When proposing the plantar pressure acquisition system,this article first selected the Flexiforce A401 membrane pressure sensor produced by the American Tekscan company,and then we studied the plantar pressure distribution,and obtained three information containing abundant plantar pressure.Position to place the pressure sensor.Then the circuit realization of the pressure sensor is studied.After solving the method of collecting human body dynamics information,this paper uses the inertial measurement unit(IMU)produced by the Dutch company Xesens to collect the kinematics information of the human lower limbs.This paper conducted a hardware system stability test on the self-built human lower limb motion data sensing collection system.The test showed that the largest signal difference ratios under three different ground environments were 6.6%,4.5%,and 7.9%,and they worked for a long time.The situation is stable,with only small standard deviation fluctuations.Then we conducted experiments on eleven independent gait modes and five conversion gait modes.The test mainly obtains Euler angles that reflect the changes in the spatial position of the lower limbs,the corresponding acceleration signals,and the plantar pressure signals of the feet that reflect the human body dynamics.After smoothing filter processing and gait segmentation processing,this paper uses the analysis window to slide through a complete gait cycle through the two-foot grounding phase and the single-foot grounding phase to extract the feature values,and then input the feature values into PCA,LDA and Three QDA classifiers with similar characteristics and structures achieved average recognition accuracy rates of 84.33%,90.93%,and 79.66% respectively.Among them,linear discriminant analysis(LDA)achieved good recognition results.In this paper,by using an autonomously constructed human lower limb movement data sensing collection system,a variety of human lower limb movement data was collected under laboratory conditions,and the LDA classifier was used to identify the human lower limb gait after extracting the feature values,and finally got the independent gait The average recognition accuracy rate of 90.93% and the average recognition accuracy rate of 92.76% in the converted gait show that the gait prediction is highly consistent with the actual movement trend,which can provide a reference for the gait recognition,prediction and control of the prosthesis.