Center of Mass Estimation for Balance Evaluation Using Convolutional Neural Networks

Abstract

The Center of Mass (CoM) position of the human body is an important indicator when evaluating a person’s balance ability. Traditionally the CoM position is measured using laboratory-grade devices like a force plate, which is expensive and inconvenient for home use. In this paper, we propose a deep learning-based framework that uses a single depth camera to estimate the CoM position of a human subject. The proposed framework takes the depth image captured by the depth camera as input, and uses supervised learning to estimate the subject’s horizontal CoM position. The model is trained and tested on data collected from multiple subjects in various postures. Evaluation results demonstrate the high accuracy of the proposed approach in estimating the CoM of existing subjects or a new subject. Compared with existing CoM estimation techniques, the proposed framework is easy to set up and does not need any subject identification process, which makes it convenient for home use. The proposed framework can be used as a portable and low-cost tool for CoM measurements and can enable automated balance evaluation at home.