Design Of Human Lower Limb Motion Data Acquisition System And Research On Motion Intention Recognition Model

The lower limb exoskeleton is an assisted walking device that combines mechanical,sensing,and control technologies and has been widely used in military,medical,and industrial fields.The key problem of lower limb exoskeleton research is to achieve precise perception of the wearer’s motion intention by the exoskeleton,so as to achieve coordination of human-machine motion.Considering that a single signal is not enough to judge all the information of gait patterns.For this reason,this project builds a lower limb motion information acquisition system,obtains the multi-sensor information of lower limb motion,and uses the gait characteristics of lower limb motion extracted by sliding window and principal component analysis(PCA),and finally identifies and predicts various motion patterns and gait phases of lower limbs by support vector machine(SVM)algorithm.The main contents of this paper are as follows:(1)A human lower limb motion data acquisition system was built,including a plantar pressure module,a muscle stiffness sensor,a joint angle module and a data relay station.The plantar pressure module consists of a plantar pressure insole and a data acquisition module,and the plantar pressure insole consists of a homemade flexible thin-film pressure sensor and a flexible printed circuit(FPC).The homemade flexible thin-film pressure sensor is made from a polydimethylsiloxane(PDMS)/carbon black composite.In order to determine the optimal carbon black doping ratio,experiments were designed for the composite carbon black doping ratio,and the optimal carbon black mass percentage obtained experimentally was 5～6wt%.The data acquisition modules were designed as a Zig Bee-based wireless acquisition module and an STM32-based wired acquisition module,respectively.The muscle hardness module uses the muscle hardness sensor to sense the degree of muscle force generation and the structure design of the muscle hardness sensor.The joint angle module uses a magnetic angle sensor made by AS5600 chip.The data relay station is able to collect information from all modules and send the collated information to the computer in a unified manner.(2)The gait phases were divided into flat walking and continuous up the ladder movement with the boundaries of heel touch,full foot touch,heel off the ground,and toe off the ground,and the continuous deep squatting up was divided into four phases: sitting zone,rising zone,standing zone,and deep squatting zone.The comparison experiments with EMG signals showed that the muscle hardness information collected by the muscle hardness module could reflect the muscle force generation.In order to determine the best binding position of the muscle hardness sensor,a comparison experiment of the binding position of the muscle hardness sensor was conducted.The results showed that the best binding position was in the middle.In addition,a correlation experiment between the signal collected by the muscle hardness sensor and the signal collected by the EMG signal sensor was conducted.The experimental results showed that the signals collected by the muscle stiffness module could accurately reflect the force output of the muscle.(3)The gait information is sampled using the sliding window technique and the time and frequency domain features are extracted to obtain 102-dimensional features.Then the10-dimensional features with the highest weights are extracted using the PCA algorithm as the input to the support vector machine(SVM)model.The model results show that the human motion intention recognition model can well recognize the motion patterns of walking on a flat road,continuous up a ladder and continuous deep squatting up,achieving 90.62%,96.15% and 90.62% accuracy,respectively.In addition,the recognition accuracy of human gait phase is also very high,with 99.13%,98.26% and98.36% for flat walking,continuous stair climbing and continuous deep squatting movements,respectively.