Locomotion Mode Recognition based on Foot posture and Ground Reaction Force

Efficient and accurate locomotion mode recognition is the basis and the key to compliance control of exoskeleton. Considering the fact that single ground reaction force information can not realize continuous gait phase recognition in the whole gait cycle and traditional inertial sensors' installation method tends to cause binding error, this paper established a wearable gait analysis system based on inertial sensor and foot pressure sensor which entirely embedded in a shoe insole. An inertial sensor mounting structure for foot posture information acquisition was designed to reduces the binding errors caused by the uncertain binding position and the random errors caused by the relative position changes with the movement. Two force sensors was set to measure the force load on the shoe insole at the heel and the forefoot during walking. Then, the gait curve of the normal human beings measured by this wearable gait analysis system is segmented by a periodic segmentation method combining power spectrum and feature points, and features is extracted from the time series of sensor signals according to the characteristics of the human gait. Probabilistic neural network is used to identify the locomotion modes and to verify the effectiveness of this wearable gait analysis system, experiments on different terrains are performed. The experiment results show that this method can effectively reduce binding error and random error, and reflect the foot movement during walking. Furthermore, the measurement method can accurately and effectively identify level-ground walking, stair ascent and stair decent, showing great potential for further development and applicability in control of exoskeleton.