CNN-LSTM-based motion phase recognition for hip exoskeleton

Abstract

Recent advancements in robotic exoskeletons have highlighted the importance of gait phase detection for enhancing mobility. This process, crucial for proper ankle function, increasingly relies on inertial measurement units (IMUs) due to their ability to provide data on angular velocity and acceleration. Traditional methods like Hidden Markov Models (HMMs) face challenges in capturing long-term dependencies. To address this, our study proposes an improved gait phase detection algorithm that combines a convolutional neural network (CNN) with a long- and short-term memory (LSTM) network, effectively integrating prior values to better understand gait dynamics. Data from seven participants were collected using an optimized Harris Hawk Algorithm to train and optimize a Support Vector Machine (SVM) device attached to their feet. The model's performance was evaluated using accuracy, precision, recall, and F1 score, achieving an F1 score of 93.33%. This algorithmic classifier, adaptable to various gait kinematics with a single foot sensor, advances gait analysis, improving control and adaptation of assistive devices and enhancing rehabilitation therapy and prosthetic development.