3D Pose Based Feedback for Physical Exercises

TL;DR

This paper presents a data-driven, learning-based framework using Graph Convolutional Networks to identify and correct mistakes in physical exercises, enhancing safety and personalization in self-rehabilitation and training.

Contribution

It introduces a novel GCN-based approach that learns to identify and correct exercise mistakes without relying on hard-coded rules, adaptable to individual users.

Findings

Achieves 90.9% mistake identification accuracy

Corrects 94.2% of identified mistakes

Introduces a new dataset with 3 exercises

Abstract

Unsupervised self-rehabilitation exercises and physical training can cause serious injuries if performed incorrectly. We introduce a learning-based framework that identifies the mistakes made by a user and proposes corrective measures for easier and safer individual training. Our framework does not rely on hard-coded, heuristic rules. Instead, it learns them from data, which facilitates its adaptation to specific user needs. To this end, we use a Graph Convolutional Network (GCN) architecture acting on the user's pose sequence to model the relationship between the body joints trajectories. To evaluate our approach, we introduce a dataset with 3 different physical exercises. Our approach yields 90.9% mistake identification accuracy and successfully corrects 94.2% of the mistakes.