Continuous Human Gait Phase Estimation Considering Individual Diversity

Accurate identification of human motion state is of great significance for exoskeleton robots to provide appropriate auxiliary torque. Previous work generated exoskeleton assist torque by recognizing the phase of the human gait. However, they set the gait phase to increase linearity throughout the gait cycle and then reset to 0 at the end, without taking into account the diversity of human gait. Different individuals exhibit varying gait characteristics, even the same person may walk with different gaits under distinct situations, such as speed and slope. To address this issue, we propose a novel gait phase coding method based on dynamic time warping (DTW), which can generate differentiated phases according to different gaits. Additionally, we introduce a long short-term memory (LSTM)-based gait phase estimation method. Our experimental results demonstrate that the proposed method achieves satisfactory performance.