Open AccessAZURE KINECT BODY TRACKING UNDER REVIEW FOR THE SPECIFIC CASE OF UPPER LIMB EXERCISES
Author(s) -
Eugenio Ivorra,
Mario Ortega,
Mariano Alcañíz
Publication year - 2021
Publication title -
mm science journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.195
H-Index - 10
eISSN - 1805-0476
pISSN - 1803-1269
DOI - 10.17973/mmsj.2021_6_2021012
Subject(s) - reliability (semiconductor) , computer science , tracking (education) , artificial intelligence , pearson product moment correlation coefficient , correlation coefficient , computer vision , lower limb , motion (physics) , match moving , motion analysis , mathematics , statistics , machine learning , medicine , surgery , psychology , pedagogy , power (physics) , physics , quantum mechanics
A tool for human pose estimation and quantification using consumer-level equipment is a long-pursued objective. Many studies have employed the Microsoft Kinect v2 depth camera but with recent release of the new Kinect Azure a revision is required. This work researches the specific case of estimating the range of motion in five upper limb exercises using four different pose estimation methods. These exercises were recorded with the Kinect Azure camera and assessed with the OptiTrack motion tracking system as baseline. The statistical analysis consisted of evaluation of intra-rater reliability with intra-class correlation, the Pearson correlation coefficient and Bland–Altman statistical procedure. The modified version of the OpenPose algorithm with the post-processing algorithm PoseFix had excellent reliability with most intra-class correlations being over 0.75. The Azure body tracking algorithm had intermediate results. The results obtained justify clinicians employing these methods, as quick and low-cost simple tools, to assess upper limb angles.