A Deep Learning-Based Approach for the Classification of Gait Dynamics in Subjects with a Neurodegenerative Disease

Neurodegenerative diseases cause changes in neuromuscular tissues through a deterioration of the motor neurons, which make the motor capability of a patient increasingly abnormal. In particular, walking is one of the movements most significantly influenced by the deterioration process. An early detection of emerging anomalies in the walking patterns of elderly subjects may help to prevent connected risks. The current walking patterns assessment methods are generally performed in supervised clinical environments and show limitations in terms of cost and accuracy. In this work, we aim to provide a contribution to the analysis of walking patterns so we address the problem of the recognition of gait dynamics by the exploration of the application of deep learning algorithms. In order to prove the goodness of our work, we have carried out five experiments, each with a different classification task. The results achieve a classification accuracy which is better by 3.9% than the accuracy achieved by models presented in related works.