Prediction of Optimal Walking Speed for Powered and Passive Prostheses to Improve long-term Gait Stability

This study explores the effects of powered versus passive prosthetic knees on gait biomechanics in transfemoral amputees with unilateral left leg amputation. Knee joint Kinetics during gait viz., Joint moment, Joint power and Ground reaction forces of the affected and unaffected legs were analyzed. Significant variations in hip joint activity were noted, highlighting compensatory adjustments in the intact limb. For the choice of prosthetics and optimal walking speeds (slow, medium and fast), machine learning techniques such as Long Short-Term Memory (LSTM) networks, Random Forests (RF), and Support Vector Machines (SVM) were employed and the results were compared. The LSTM model achieved an accuracy of 97% in predicting prosthetic type and walking speed, surpassing the 94% accuracy of SVM and the 95% accuracy of RF models. These results underscore the importance of personalized prosthetic solutions to improve gait efficiency and reduce compensatory movements, thereby enhancing long-term mobility and comfort.