RGB-D ergonomic assessment system of adopted working postures

An Evaluation of Posture Recognition Based on Intelligent Rapid Entire Body Assessment System for Determining Musculoskeletal Disorders

“…For the scores, 83.33% of the scores were consistent with those of the experts. P 0 ranged from 0.968 to 0.931 and kappa ranged from 0.763 to 0.710, which were higher than those obtained by Abobakr et al [39] (with an average P 0 of 0.846 and ranged from 0.86 to 0.82). The average value of kappa was 0.700 and ranged from 0.78 to 0.63, while Plantard et al [47] obtained P 0 values ranging from 0.77 to 0.62 with an average of 0.712 and kappa values ranging from 0.66 to 0.46, with an average of 0.624.…”

“…By using the motion capture system and utilizing experts as benchmarks, the reliability and feasibility of Quick Capture were verified. The results were compared to that of Abobakr et al [39] and Plantard et al [47]. It was proven that the Quick Capture system was slightly better than their results in terms of angle recognition, and its scores were higher than theirs.…”

“…Wei et al [38] developed a sequential architecture consisting of a convolution network for posture recognition and integrated convolutional pose machines (CPM) for joint posture estimation. Abobakr et al [39] proposed a deep convolutional neural network to predict body joint points from a single image and used angle data for MSDs risk assessment using the RULA method. From the experimental comparison between the risk level score from experts and the abovementioned neural network, the prediction accuracy obtained was 89%, the kappa index was 0.71, and the average mean absolute error (MAE) was 3.19 • ± 1.57 • .…”

Prototype System for Measuring and Analyzing Movements of the Upper Limb for the Detection of Occupational Hazards

“…Observation-based methods range from checklists and diagrams using software [ 15 , 16 , 17 ], questionnaires for quantitative ergonomic analysis [ 18 ], to video observation or direct measurement [ 18 ]. Some of these methods, such as those based on observational measurements of posture from videos, have been studied and offer great confidence in the results obtained [ 19 ]. Other studies propose methods that analyze the articulated posture, which predicts the angles of the body’s joints from a single depth image [ 19 ], or that take quick snapshots with a smartphone to recognize angles, thus analyzing users’ positions [ 20 ].…”

“…Some of these methods, such as those based on observational measurements of posture from videos, have been studied and offer great confidence in the results obtained [ 19 ]. Other studies propose methods that analyze the articulated posture, which predicts the angles of the body’s joints from a single depth image [ 19 ], or that take quick snapshots with a smartphone to recognize angles, thus analyzing users’ positions [ 20 ]. In this line, but with a system of pressure sensors, a classification of human standing posture from standing-pressure images is proposed, by means of which a robot system can predict the foreseen actions of human workers in a HRC environment (Human Robot Collaboration) [ 21 ].…”