High Real-time Locomotion Recognition Algorithm Using Inertial Measurement Units

Human-machine interaction devices (HMI), such as assisted exoskeletons and assisted prostheses, that improve human motion behaviours are playing an increasingly important role in various fields. In order for these HMI devices to provide accurate and appropriate assistance to wearers, the real-time recognition of locomotion with accuracy is among the key issues to be solved. This paper proposes a high real-time motion pattern recognition algorithm using inertial measurement units (IMU) that can accurately and quickly identify the current motion state by collecting various motion parameters of the lower limb during human walking. The algorithm uses a trajectory solving and motion direction discrimination method based on the shank zero velocity point and a terrain classification method based on an F-score improved support vector machine (F-score-SVM). In this paper, experimental data was used to verify the performance of the algorithm. Results show that the algorithm performs better than traditional recognition algorithms in complex environments.