Intelligent Human Gait Phase Classification Using Machine Learning Models

Advancements in machine learning (ML) have facilitated the prediction of key aspects of human locomotion, particularly in identifying subject gait trajectories essential for recognizing neurological and pathological gait patterns. In this study, data from 40 healthy people were used to predict gait phases using lower limb joint angles as the inputs. The gait cycle’s main and sub-phases were predicted using Artificial Neural Networks (ANN), Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Ensemble and Phase-aware Attention-based Convolutional and Pooling Deep Neural Network (pACP-DNN) learning techniques. A comprehensive evaluation was performed using 5-fold and 10-fold cross-validation to verify model robustness. The proposed approach achieved an accuracy of 0.99 for main phase prediction and 0.94 for sub-phases, exceeding previous research and confirming its efficacy in gait phase categorization. These well-performing models hold significant promise for contributing to the development of assistive equipment for individuals with physical disabilities.