Locomotion Mode Recognition Based On Multiple Sensors

The Locomotion mode is a classification of human daily life behaviors.Motions such as walking,running,stair ascent and stair ascent are different locomotion modes.Lower extremity exoskeleton system is designed to follow the movement of human lower limb in multiple motion scenes.The human lower limb joint angles vary from mode to mode.Thus,recognition of the current locomotion mode is required to provide key control input for the system.Acceleration and plantar pressure data contain a wealth information of human gait and movement,which can help determine the person's motion states.For the purpose of the identification of different locomotion patterns,three parts will be introduced in this article,including system software and hardware,key gait events identification and locomotion mode recognition.First of all,a system of data measurement and acquisition is designed based on the characteristics of human locomotion.The system consists of five accelerometers and a pair of pressure measurement shoes,which can collect acceleration data from multiple site and plantar pressure data from shoes simultaneously.All data will be transmitted or stored after acquired for several kinds of purpose.Installation method,battery life,and device size are adequately considered to meet the need for continuous collection and test experiments outdoors in this design.Then,the plantar pressure data was collected under different motion modes,and the dynamic gait monitoring was used to identify key gait events including toe off and heel contact by properly designing the experiment.For the misjudgment of key gait events in the monitoring process,a masked window method is purposed to filter the repeatedly identified events in a gait cycle,so as to improve the recognition accuracy and provide gait information for the recognition of the locomotion mode.Finally,for the problem that the sliding window may result in a low recognition rate of the motion pattern,different classifiers are built on the acceleration data before and after the key gait events to capture the acceleration data,and the characteristics of acceleration in each time window are extracted.The current locomotion mode status is discriminated and determined based on the output of multiple classifier.The experimental results show that the average accuracy rate of recognition can reach 95.84% when the method purposed only extracts simple statistical features.