The Research Of Gait Pattern And Gait Transition Based On The Motion Information

Recent years, due to the real time communicate with exoskeleton wearer, the lower exoskeleton has become a hot topic in the field of robotics. When people wear the lower exoskeleton, the exoskeleton can combine with wearer organically and has the function of protecting and supporting. While the wearer also can expand the limb capacity and improve muscle endurance, even complete the task which can’t be completed under natural conditions. Based on the above advantages, the lower exoskeleton has broad application prospects in both military and medical rehabilitation fields.After summarizing the research status and research significance of the lower exoskeleton, the gait is studied in detail in this thesis. The content mainly involve:the acquisition and processing of the motion information, the recognition and analysis of the gait pattern, the analysis and planning of the gait transition process.The first problem needed to be solved of the lower exoskeleton is how to acquire and appropriately process the wearers’ motion information. The lower limb motion information, including the inertial data of the thighs and calves and the sEMG of the key muscles, is extracted through the micro inertial sensors and the sEMG acquisition system in this thesis Two types of the motion features are extracted by wavelet decomposition, the time-frequency domain analys and other methods. The BPSO algorithm is applied to fuse the motion features and the fusion features which have moderate dimensions to classify the different gait patterns effectively are obtained. The SVM classifier is employed to compare the results of the classification and identification of different types of features and verify the validity of the algorithm.After acquiring the wears’ motion information, the control strategy is needed to adjust exoskeleton itself according to the identified wearers’gait patterns and ensure the coordinated movement of the exoskeleton and its wearer. The combinatorial classification model of the HMM and SVM which comprises the advantages of HMM that is better at modeling the changing process and SVM that is better at recognizing the discrepancies between classes is proposed in this thesis. It is used to classify and recognize five common gaits including walking on the flat road, going up the stairs, going down the stairs, uphill walking and downhill walking. The superiority of the algorithm is verified by comparing the results of classification.The gait transition process is generally involved in the motion process. The gait transition processes of "walking on the flat road transition to going up and down the stairs" and "walking on the flat road transition to uphill and downhill walking" are researched in this thesis. The application value of sEMG in the recognition of gait transition is proved by analyzing the characteristics of sEMG during the gait transition processes. The HMM based on the fusion features of the motion information is used to recognize the intention of the gait transition. Meanwhile, the curves of the joint angles are planned by the interpolation method during the gait transition process of "walking on the flat road transition to going up the stairs".