Research On Gait Recognition Algorithm Of Lower Extremity Exoskeleton Robot Based On Multi-sensor Fusion Data Acquisition System

In recent years,the lower limb exoskeleton robot has been widely studied and applied in various fields,especially in the military field,the wearing of the lower limb exoskeleton robot can help soldiers to bear weight,and in the medical field,the assisting lower limb exoskeleton robot can help the disabled walk or carry out rehabilitation training.The fast,high-precision and multi classification gait recognition method has a key impact on the precise control of the lower extremity exoskeleton robot.Therefore,the main work of gait recognition under the condition of wearable exoskeleton robot is as follows:(1)Through the research on the control principle of motor-driven and electric cylinder driven power exoskeleton robot,it can be seen that the main control methods are position control and motion attitude switching.Therefore,The main contents of gait recognition in this thesis are gait phase recognition and motion attitude recognition.Based on the basic analysis of human walking gait phase and motion posture,the important influence of joint angle and foot pressure on gait phase recognition and motion posture recognition is revealed.(2)Aiming at the data acquisition of joint angle and plantar pressure,a gait data acquisition system based on inertial sensor and membrane pressure sensor is designed.And the upper computer interface is designed by Lab VIEW,which realizes the real-time storage and real-time monitoring of gait data,and provides the data acquisition conditions for subsequent experiments.(3)In the aspect of gait phase recognition,a fuzzy recognition algorithm is proposed based on the membership function,fuzzy rule and anti fuzzy method.Experiments show that the algorithm has a good recognition rate and can provide data support for the position control of exoskeleton robot.(4)In the aspect of motion attitude recognition,based on the support vector machine theory,a motion attitude recognition algorithm with RBF radial basis function as the kernel function is designed.The time-domain features,FFT coefficients and self-determined features are selected to form the feature set,which is brought into the recognition algorithm to train the motion recognition model.In order to avoid over fitting and improve the recognition rate,an improved SVM-RFE recognition algorithm is proposed.The feature set is selected by the recursive feature elimination method,and the optimal feature subset is obtained.The optimal motion recognition model is obtained by training feature subset.Experimental results show that the improved algorithm has better recognition accuracy and recognition effect than the general algorithm,which can provide theoretical ideas for the motion attitude switching of exoskeleton robot.