A Novel Non-contact Recognition Approach of Walking Intention Based on Long Short-Term Memory Network

To daily assisted walking and walking rehabilitation training scenarios, it is especially important that the robot accurately recognizes user’s walking intentions. To accurately identify walking intention in the process of operating walking rehabilitation training robot, a novel non-contact walking intention recognition method based on LSTM (Long and Short-Term Memory network) is proposed under this paper. Firstly, this paper introduces the mechanical structure of walking rehabilitation training robot and establishes the kinematics model of the robot. Secondly, the distance information of the left and right legs is detected by a multi-channel proximity sensor, and speed information of the left and right legs is obtained using a distance-speed conversion algorithm. The distance information of the left and right legs and speed information of the left and right legs is used as the input of the LSTM algorithm. The multivariable LSTM is utilized to predict the user’s walking intention to obtain the desired movement speed of the robot. Finally, the algorithm is tested for walking intention recognition experiment. The experiment shows that the algorithm is suitable for diverse users. Users walk at constant speed and variable speed walking, walking rehabilitation training robot has a speedy recognition and reaction ability. The multi-channel proximity sensor used in this paper can detect information in a non-contact manner, which provides users with a comfortable and unconstrained walking experience.