A Non-contact Method in Estimating Energy Expenditure in Aerobics Based on OpenPose Algorithm

Objective A non-contact method estimating the energy expenditure of aerobics was established and verified based on OpenPose algorithm.

Methods The validity of joint angle recognition using OpenPose algorithm was determined with Qualisys motion capture system. Using the energy expenditure measured by COSMED K5 wearable metabolic system as criteria, a novel method based on OpenPose algorithm was established and validated to estimate the energy expenditure of aerobics.

Result Most joint angles in human movement could be accurately measured by OpenPose algorithm. The error of energy consumption estimation method based on OpenPose computer vision algorithm was fairly good (MAE=0.81, MSE=1.11, RMSE=1.02, MRE=13.60%), and the accuracy could reach 86.4%.

Conclusion A non-contact measurement method of energy consumption in aerobics based on OpenPose algorithm was thus initially established, which could measure the energy consumption in aerobics both accurately and easily.