Convolutional neural network model for gait classification of flexible lower limb exoskeleton

Gait refers to the way humans walk. Gait is a complex behavioral characteristic. It is very challenging to identify the intention of human lower limbs for lower limb exoskeleton robots. How to improve the accuracy of gait classification tasks under the premise of reducing complexity is our pursuit. In this paper, a convolutional neural network-based method is proposed to recognize and classify human motion states in outdoor activities while wearing a flexible lower limb exoskeleton. This method hopes that the exoskeleton can meet the tasks of convenient wearing and high accuracy of gait recognition for the elderly, and realize their good wishes of improving the quality of life and increasing the time of outdoor activities. The inertial measurement unit mounted on the hip joint of the lower limb collects motion data under different gaits, Including Euler Angle data, angular velocity data, acceleration data, and magnetic field strength data. After data collection and processing, a python script is used to segment the data into a certain size matrix, and then as the input of a convolutional neural network; various gait activity types are classified and recognized. By using this method, the recognition rate of wearers in different terrain can reach 98.6%, indicating that this method can ensure the good effect of the exoskeleton in gait activity recognition of human lower limbs