Analysis of Motion Intention Prediction Based on Human Gait Data

In the research of lower limb rehabilitation robotics, accurate prediction of human motion intention plays a key role in achieving active robot control, providing personalised rehabilitation programmes and enhancing training effects. Although there have been more studies on motion intention recognition methods based on physiological signals (e.g. EMG and EEG signals), there is still a relative lack of research on continuous motion estimation at multiple joint angles. To address this, a multi-objective gait prediction model is developed to simultaneously estimate the motion trajectories of the bilateral hip and knee joints. This model integrates joint position data obtained from the NOKOV motion capture system with a hybrid architecture combining convolutional neural networks (CNN) and long short-term memory (LSTM) networks. Experimental results demonstrate the model's capability to accurately predict joint motion trajectories, providing robust data support for the active control of lower limb rehabilitation robots, and highlighting its potential for practical implementation in rehabilitation training protocols.