Prediction of characteristic points of hip and knee joint trajectories during overground walking using IMUs and Artificial Neural Networks

This paper presents a study on overground non-pathological gait, focusing on hip and knee joint trajectories in sagittal plane. The objects of study are some characteristic points of the joint curves (including the extrema) and their relation to gait parameters, namely normalized walking speed, cadence and normalized step length. The main objective is to predict the spatio-temporal values of these points depending on given gait parameters. To this end, a study with 18 healthy subjects was conducted, where they were asked to walk as comfortable as possible whilst following different tasks, namely walking with desired and given cadence, step length and speed. The data was processed and fed to artificial neural networks to obtain an algorithm able to predict the characteristic points. Specifics of the study protocol and data processing are presented, as well as the prediction results.