Research And Implementation Of The Exoskeleton Gait Detection System

With the exoskeleton robot technology continues to evolve, its fast, effective andstable control attracts more and more people’s attention. To meet the controlrequirement of external skeletal robot, it needs to accurately detect information gait,while the degree of detection accuracy will directly affect the entire outer skeletalsystem control effect is good or bad, so designing and building a high detectionaccuracy of the step state detection system is particularly important.The widely used methods for the detection of gait depends on human experience,gait detection information used in the process is relatively simple, and the detectionmethods used in fixed without dynamic updates, these will affected gait detectionaccuracy. Currently in gait detection system for problems, we designed andimplemented a new gait detection system. The system of data acquisition, preprocessing,training and the establishment of basic classification model, and then based on thismodel to identify classifications, while the recognition result to join an existing databaseclassification model of the dynamic update and the updated model for the next momentin which the detection of gait.By studying and analyzing gait detection systems currently used in the gait cycleclassification, the paper ultimately select the most human gait kinematics of eight phaseclassification, which also got a precise identification of the eight categories. With sucheight completed two Phase Classification dynamic classification and identificationmethod is implemented——Dynamic classification Euclidean distance and SupportVector Machine (SVM) dynamic classification method, their giat detection method ismore innovatively in the algorithm while still on the increase in the number, the methodusing a single current detection method of the pressure information, acceleration sensorinformation together, increases the amount in real time to ensure the system controlvalid under the premise of improving the detection accuracy gait.In this paper, self-validation, Leave one out cross-validation, Test on other subjectsas well as other experimental comparison, respectively, the method repeatability, reliability, and other aspects of characteristics to be verified, and has made moretraditional detection methods higher detection accuracy rate and better detection effect,in order to better meet the exoskeleton system testing and control requirements.Meanwhile, the method for some unexpected abnormal signal points generated aconsiderable inhibition eliminate the role, the class of the signal shows a goodrobustness properties.