| For a long time,diseases,natural disasters and accidental injuries have brought tens of millions of patients with physical disabilities to this society.In recent years,the number of patients with physical disabilities has increased rapidly.Data from the second national sample survey of disabled people shows that China now has 24.12 millions people with physical disabilities and 440 thousands of lower limb amputees.Therefore,physical disability has become one of the major disability diseases in China's population,and the society has to pay a great price to place and compensate the vulnerable groups.At present,intelligent prostheses can restore normal walking functions to amputees,but the prosthetics on the market are passive or semi-active,and cannot help walking.The lower extremity assisted exoskeleton robot can help to walk,but it is only suitable for people with healthy lower limbs.There is no corresponding product for above-knee amputees.Therefore,our research team puts forward an asymetric lower extremity exoskeleton(LEE)for above-knee amputees in order to enable the amputee walk with a load and walk easily in daily life.To enable the exoskeleton to assist above-knee amputees in the procedure of walking on flat ground or going up and down stairs,this paper proposed a movement pattern recognition and joint angle prediction method for the amputation side of LEE based on EMG signals hip joint angle and plantar pressures.The main contents of this paper include:(1)The establishment of experimental platform and analysis of EMG signal mechanism.According to the structure of the asymmetric LEE,a simplified experimental platform for the amputation side was designed.The kinematics model of the amputation side of the exoskeleton robot was established.The mechanism of EMG signal is analyzed to provide a theoretical basis for pattern recognition.(2)Online gait planning of asymmetric LEE.Firstly,thin-film pressure sensors and six-axis angle sensors were selected to build the signal acquisition system.Then the experiment platform was built.Arduino microcontroller was used for software programming,signal acquisition and processing.The walking gait was divided into the pro-support phase,mid-support phase,late-support phase and the swing phase according to the plantar pressure signals.The feasibility of online gait planning was theoretically verified by the measurement results.According to the signal measured by the angle sensors,the method suitable for curve fitting of each joint was selected.(3)Lower limb motion pattern recognition based on multi-source signals.A multi-source signal acquisition system was designed,which can synchronously measured the plantar pressure signals,sEMG signals.According to the plantar pressures and the EMG signals of three positions,the patterns of walking up stairs,walking down stairs,walking on flat ground,sitting up and standing were recognized.At the end of the support period,the normal swing and stop walking patterns were also recognized according to the EMG signals using PSO-SVM.(4)Joint angle acquisition,co-simulation and experiment under different movement patterns.The curves of joint angle under different motion modes were measured and fitted using polynomial A method of mapping the knee joint and ankle joint angles from the hip joint angle was proposed.Simulink/Adams simulation platform was also built to verify the correctness of the pattern recognition system and the feasibility of joint angle mapping method.(5)Joint angle prediction in walking up and down stairs based on GRNN.According to the process of walking upstairs and downstairs,three lower limb joint angles were predicted by EMG signals,plantar pressure values and hip joint angle.Generalized regression neural network(GRNN)was used for mapping and the predicted results were analyzed. |
PDF Full Text Request