Gait Recognition Method Based On Attitude And Pressure Information

In order to achieve the exoskeleton robot’s fast, effective and stable control, one needs to identify the gait information. Whether the recognition is fast and accurate will directly affect the whole exoskeleton robot control system. Therefore, designing an accurate gait recognition system is of great significance to the exoskeleton robot control. Currently the threshold value method depending on experience is widely used, but it only detects the trademark in walking process, cannot achieve all phase identification continuously. The detected information is relatively simple and cannot be truthful to show the human gait.Human gait is changing while walking because of different objects and different walking speed, and each gait phase also has certain duration and unclear boundaries. This thesis puts forward a gait recognition method fusing the angle signals of each lower extremity joints and plantar distributed pressure by using the fuzzy theory. Through multiple sets of sample data, and using comparative analysis, it is proved that this method has high accuracy in walking mode under different speed and height, and is able to approach quick and efficient gait recognition with changing speed.Main work of the thesis is as follows:firstly, gait phases are divided by analyzing the movement mechanism of human lower extremity. Secondly, the sensing system which detects lower extremity movements and distributed foot pressure is designed to get human movement information, and filtering algorithm is used to process the motion information. Again, the multi-sensor information fusion process and different levels are researched, and the focal point of above is multi-sensor fusion recognition algorithm based on fuzzy theory. A gait recognition algorithm is provided, in which each sensor’s information is fused at feature level, and then the final recognition result is obtained by global fusion with all sensors. Finally, gait information detecting experiment is designed to analyze amplitude, peak and cycle of each posture data. Through multiple experiments in four different speed and four different height profiles, analysis results prove feasibility and effectiveness of this method.