Prediction accuracy in estimating joint angle trajectories using a video posture coding method for sagittal lifting tasks

This study investigated prediction accuracy of a video posture coding method for lifting joint trajectory estimation. From three filming angles, the coder selected four key snapshots, identified joint angles and then a prediction program estimated the joint trajectories over the course of a lift. Results revealed a limited range of differences of joint angles (elbow, shoulder, hip, knee, ankle) between the manual coding method and the electromagnetic motion tracking system approach. Lifting range significantly affected estimate accuracy for all joints and camcorder filming angle had a significant effect on all joints but the hip. Joint trajectory predictions were more accurate for knuckle-to-shoulder lifts than for floor-to-shoulder or floor-to-knuckle lifts with average root mean square errors (RMSE) of 8.65°, 11.15° and 11.93°, respectively. Accuracy was also greater for the filming angles orthogonal to the participant's sagittal plane (RMSE = 9.97°) as compared to filming angles of 45° (RMSE = 11.01°) or 135° (10.71°). The effects of lifting speed and loading conditions were minimal. To further increase prediction accuracy, improved prediction algorithms and/or better posture matching methods should be investigated.

Statement of Relevance: Observation and classification of postures are common steps in risk assessment of manual materials handling tasks. The ability to accurately predict lifting patterns through video coding can provide ergonomists with greater resolution in characterising or assessing the lifting tasks than evaluation based solely on sampling with a single lifting posture event.     

Estimation of compressive forces on lumbar spine from categorical posture data

To combine estimates of trunk posture and force into an integrated measure of load on the low back, continuous variables for body angles were estimated by assuming specified distributions within corresponding posture categories with Monte-Carlo (MC) simulation. The estimated posture angles were compared with reference measurements from the Lumbar Motion Monitor and inclinometers. The lumbar compression estimates, generated from simulated posture angles and from direct measurement, were compared. Trunk flexion showed high correlation between direct measurements and simulated angles, as did L5/S1 compression. The MC approach to extracting continuous posture angles from categorized observations did not appear to introduce large error in the variables used to estimate spinal compressive forces. When instrumentation methods of postural assessment are not feasible, a simulation approach combined with biomechanical modelling could be used to integrate multiple external exposure variables into estimates of compressive forces acting on the low back.     

Influence of fatigue in neuromuscular control of spinal stability

Lifting-induced fatigue may influence neuromuscular control of spinal stability. Stability is primarily controlled by muscle recruitment, active muscle stiffness, and reflex response. Fatigue has been observed to affect each of these neuromuscular parameters and may therefore affect spinal stability. A biomechanical model of spinal stability was implemented to evaluate the effects of fatigue on spinal stability. The model included a 6-degree-of-freedom representation of the spine controlled by 12 deformable muscles from which muscle recruitment was determined to simultaneously achieve equilibrium and stability. Fatigue-induced reduction in active muscle stiffness necessitated increased antagonistic cocontraction to maintain stability resulting in increased spinal compression with fatigue. Fatigue-induced reduction in force-generating capacity limited the feasible set of muscle recruitment patterns, thereby restricting the estimated stability of the spine. Electromyographic and trunk kinematics from 21 healthy participants were recorded during sudden-load trials in fatigued and unfatigued states. Empirical data supported the model predictions, demonstrating increased antagonistic cocontraction during fatigued exertions. Results suggest that biomechanical factors including spinal load and stability should be considered when performing ergonomic assessments of fatiguing lifting tasks. Potential applications of this research include a biomechanical tool for the design of administrative ergonomic controls in manual materials handling industries.     

Humanoid Posture Selection for Reaching Motion and a Cooperative Balancing Controller

Our goal in this research was to develop a motion planning algorithm for a humanoid to enable it to remove an object that is blocking its path. To remove an object in its path, a humanoid must be able to reach it. Simply stretching its arms, which in a humanoid are shorter than its body and legs, is not sufficient to reach an object located at some distance away or on the ground. Therefore, reachability has to be ensured by a combination of motions that include kneeling and orienting the pelvis. However, many posture selection options exist because of the redundancy of a humanoid. In this research, we focused on the optimization of the posture of a humanoid that is reaching toward a point. The posture selected depends on the initial posture, the location of the point, and the desired manipulability of the humanoid’s arms. A cooperative balancing controller ensures the stability of the reaching motion. In this paper, we propose an algorithm for reaching posture selection and a balancing controller for humanoids, and we present the results of several experiments that confirm the effectiveness of the proposed algorithm and controller.     

Posture/Walking Control for Humanoid Robot Based on Kinematic Resolution of CoM Jacobian With Embedded Motion

This paper proposes the walking pattern generation method, the kinematic resolution method of center of mass (CoM) Jacobian with embedded motions, and the design method of posture/walking controller for humanoid robots. First, the walking pattern is generated using the simplified model for bipedal robot. Second, the kinematic resolution of CoM Jacobian with embedded motions makes a humanoid robot balanced automatically during movement of all other limbs. Actually, it offers an ability of whole body coordination to humanoid robot. Third, the posture/walking controller is completed by adding the CoM controller minus the zero moment point controller to the suggested kinematic resolution method. We prove that the proposed posture/walking controller brings the disturbance input-to-state stability for the simplified bipedal walking robot model. Finally, the effectiveness of the suggested posture/walking control method is shown through experiments with regard to the arm dancing and walking of humanoid robot.     

Effect of foot movement and an elastic lumbar back support on spinal loading during free-dynamic symmetric and asymmetric lifting exertions

The aim of this study was to assess the effect of an elastic lumbar back support on spinal loading and trunk, hip and knee kinematics while allowing subjects to move their feet during lifting exertions. Predicted spinal forces and moments about the L5/S1 intervertebral disc from a three-dimensional EMG-assisted biomechanical model, trunk position, velocities and accelerations, and hip and knee angles were evaluated as a function of wearing an elastic lumbar back support, while lifting two different box weights (13.6 and 22.7 kg) from two different heights (knee and 10 cm above knee height), and from two different asymmetries at the start of the lift (sagittally symmetric and 60°asymmetry). Subjects were allowed to lift using any lifting style they preferred, and were allowed to move their feet during the lifting exertion. Wearing a lumbar back support resulted in no significant differences for any measure of spinal loading as compared with the no-back support condition. However, wearing a lumbar back support resulted in a modest but significant decrease in the maximum sagittal flexion angle (36.5 to 32.7°), as well as reduction in the sagittal trunk extension velocity (47.2 to 40.2°s^-1). Thus, the use of the elastic lumbar back support provided no protective effect regarding spinal loading when individuals were allowed to move their feet during a lifting exertion.     

Static and dynamic lifting strength at different reach distances in symmetrica] and asymmetrical planes

Postural and therefore biomechanical standardization in strength testing has not been rigorously and consistently applied. To develop a quantitative relationship between strength and posture (body position, symmetry, and reach) 30 normal subjects (18 male and 12 females) were required to stoop and squat lift or exert in the relevant posture against a standardized instrumented handle. The isometric lifting efforts and isokinetic lifts were studied. The isokinetic lifts were done at a linear velocity of 50cm/s of the hand displacement from the floor to the knuckle heights of the respective subjects in stoop and squat postures. The isometric stoop lifting efforts were exerted in two standardized postures: (a) with 60° hip flexion; and (b) with 90° hip flexion. The isometric squat lifting efforts were also exerted in two standardized postures: (a) with 90° knee flexion; and (b) with 135° knee flexion. All isometric lifting efforts and isokinetic lifts were performed at half, three-quarters, and full horizontal reach in sagitally symmetrical, 30° left lateral, and 60° left lateral planes. Isometric stoop and squat lifting efforts were also measured in self-selected optimal postures. These 56 conditions were tested in random order. The analysis of variance revealed that the gender, the mode of lifting, the postural asymmetry and reach of lifting affected the strength significantly (p<0·0001). Most two-way and three-way interactions were significant (p<0·01). Of 108 prediction regression equations, 103 were significant with up to 90% of the variation explained by anthropometric variables and sagittal plane strength. The reach affected the strength most profoundly followed by postural asymmetry and the mode of lifting.     

The effect of camera viewing angle on posture assessment repeatability and cumulative spinal loading

Video-based task analysis in the workplace is often limited by equipment location and production line arrangement, therefore making it difficult to capture the motion in a single plane. The purpose of this study was to investigate the effects of camera placement on an observer's ability to accurately assess working postures in three dimensions and the resultant influence on the reliability and repeatability of calculated cumulative loading variables. Four video cameras were placed at viewing angles of 0°, 45°, 60° and 90° to the frontal plane, enabling the simultaneous collection of views of four lifting tasks (two symmetric and two asymmetric). A total of 11 participants were trained in the use of the 3DMatch 3-D posture matching software package (developed at the University of Waterloo) and were required to analyse 16 lifting trials. Four of the participants were randomly selected to return within 72 h and repeat the analysis protocol to test intra-observer repeatability. Posture matching agreement between camera views was higher when the body segments had a minimal range of motion during the task. There was no significant participant main effect; however, there was a significant (p < 0.05) task main effect. Intraclass correlation coefficients (ICC) were calculated to assess the between day reliability. Compression, reaction anterior shear and extension moment were all found to have excellent reliability (ICC > 0.75). Joint anterior shear and joint posterior shear both provided fair to good reliability (0.4 > ICC < 0.75). Overall, the impact of the camera viewing angle on an observer's ability to match working postural exposure was found to be small.     

Estimation of compressive forces on lumbar spine from categorical posture data

To combine estimates of trunk posture and force into an integrated measure of load on the low back, continuous variables for body angles were estimated by assuming specified distributions within corresponding posture categories with Monte-Carlo (MC) simulation. The estimated posture angles were compared with reference measurements from the Lumbar Motion Monitor and inclinometers. The lumbar compression estimates, generated from simulated posture angles and from direct measurement, were compared. Trunk flexion showed high correlation between direct measurements and simulated angles, as did L5/S1 compression. The MC approach to extracting continuous posture angles from categorized observations did not appear to introduce large error in the variables used to estimate spinal compressive forces. When instrumentation methods of postural assessment are not feasible, a simulation approach combined with biomechanical modelling could be used to integrate multiple external exposure variables into estimates of compressive forces acting on the low back.     

SEE: A proactive strategy-centric and deep learning-based ergonomic risk assessment system for risky posture recognition

Work-related musculoskeletal disorders (WMSDs) are serious workplace injuries that put workers' safety at risk. However, traditional WMSD assessments are based on the human-evaluation strategy (HES), requiring human intervention. Proactive strategy (PAS)-oriented WMSDs assessments collect data using posture data tags and special semi-human–machine equipment to improve efficiency and reduce human efforts to capture specific postures in a real-world setting. Meanwhile, more research on applying artificial intelligence-based pose machines for musculoskeletal risk assessment in various workplaces is needed. Hence, this study proposed a holistic posture acquisition and ergonomic risk analysis model with the PAS-oriented philosophy for developing a smartphone-based and workplace-based risk assessment system for WMSDs. The Convolutional Pose Machines (CPM) method was combined with a rapid entire body assessment method for the system's design. Finally, the smart ergonomic explorer (SEE) system includes three subsystems: an automotive scene capturer, an ergonomic risk level calculator, and a risk assessment reporter. A musculoskeletal risk assessment experiment with 13 poses was also carried out to validate the SEE system and compare its accuracy with manual evaluation. The result shows good agreement with the REBA score, with an average proportion agreement index (P0) of 0.962 and kappa of 0.82. It indicates that the proposed system can not only accurately analyze the working posture, but also accurately evaluate the total REBA scores. This study is hoped to provide practical advice and implications for achieving a more effective empirical response for WMSD assessment.