Generative adversarial networks for generation and classification of physical rehabilitation movement episodes

TL;DR

This paper introduces a GAN-based approach for modeling, classifying, and generating human movement episodes during physical therapy, demonstrating effective motion recognition and realistic motion synthesis from recorded data.

Contribution

It presents a novel application of GANs with various architectures for modeling and classifying rehabilitation movements, validated on real motion capture data.

Findings

Networks can classify new movement instances accurately.

Models can generate realistic motion sequences similar to recorded data.

Deep subnetworks with convolutional or recurrent units are effective.

Abstract

This article proposes a method for mathematical modeling of human movements related to patient exercise episodes performed during physical therapy sessions by using artificial neural networks. The generative adversarial network structure is adopted, whereby a discriminative and a generative model are trained concurrently in an adversarial manner. Different network architectures are examined, with the discriminative and generative models structured as deep subnetworks of hidden layers comprised of convolutional or recurrent computational units. The models are validated on a data set of human movements recorded with an optical motion tracker. The results demonstrate an ability of the networks for classification of new instances of motions, and for generation of motion examples that resemble the recorded motion sequences.