The objective of this study was to develop an algorithm for automatically processing data collected with inertial measurement unit (IMU) wearable devices to measure lifting risk factors for low back disorders. Five IMU sensors attached to five body segments were used for developing the algorithm. The algorithm consists of two modules running in parallel for detecting the beginning and ending of a lifting event as well as the vertical height (V) of the object lifted by two hands and the horizontal (H) distance between the object and the body during the lift. The motion synchronization feature of wrists’ motion data were used to train the lifting detection module using a machine learning approach. This module achieved a training accuracy of 85%. In the second module, the forearm length and gyroscope data of four sensors are proposed for calculating trunk flexion angle, V and H during a lift.

The objective of this study was to assess the accuracy of an algorithm for processing data from five inertial measurement unit (IMU) sensors for measuring the vertical distance (V) and horizontal distance (H) of two handed lifting, trunk flexion angle (T) and lifting duration (LD). The sensors were placed on five body segments including the left wrist, right wrist, upper arm of the dominant hand, upper back, and thigh of the dominant leg. A laboratory-grade optical motion capture system was used as the ground truth for the assessment. Data were collected on ten subjects that performed 12 two-handed lifting tasks varying in height of the hands and horizontal distance between the body and the lifted object. Results showed that the algorithm performed well for determining the LD (~1 sec error) and T (~2° error). The average errors for V and H were about 33 and 6.5 cm, respectively.

No abstract

No abstract

Occupationally-induced back pain is a leading cause of reduced productivity in industry. Detecting when a worker is lifting incorrectly and at increased risk of back injury presents significant possible benefits. These include increased quality of life for the worker due to lower rates of back injury and fewer workers’ compensation claims and missed time for the employer. However, recognizing lifting risk provides a challenge due to typically small datasets and subtle underlying features in accelerometer and gyroscope data. A novel method to classify a lifting dataset using a 2D convolutional neural network (CNN) and no manual feature extraction is proposed in this paper; the dataset consisted of 10 subjects lifting at various relative distances from the body with 720 total trials. The proposed deep CNN displayed greater accuracy (90.6%) compared to an alternative CNN and multilayer perceptron (MLP). A deep CNN could be adapted to classify many other activities that traditionally pose greater challenges in industrial environments due to their size and complexity.

Objective A computer vision method was developed for estimating the trunk flexion angle, angular speed, and angular acceleration by extracting simple features from the moving image during lifting. Background Trunk kinematics is an important risk factor for lower back pain, but is often difficult to measure by practitioners for lifting risk assessments. Methods Mannequins representing a wide range of hand locations for different lifting postures were systematically generated using the University of Michigan 3DSSPP software. A bounding box was drawn tightly around each mannequin and regression models estimated trunk angles. The estimates were validated against human posture data for 216 lifts collected using a laboratory-grade motion capture system and synchronized video recordings. Trunk kinematics, based on bounding box dimensions drawn around the subjects in the video recordings of the lifts, were modeled for consecutive video frames. Results The mean absolute difference between predicted and motion capture measured trunk angles was 14.7°, and there was a significant linear relationship between predicted and measured trunk angles ( R2 = .80, p < .001). The training error for the kinematics model was 2.3°. Conclusion Using simple computer vision-extracted features, the bounding box method indirectly estimated trunk angle and associated kinematics, albeit with limited precision. Application This computer vision method may be implemented on handheld devices such as smartphones to facilitate automatic lifting risk assessments in the workplace.

We previously developed a method for classifying lifting postures using dimensions of a rectangular bounding box drawn tightly around the subject for a single camera view that is more tolerable of conditions encountered in industrial settings than high precision tracking and can be practically implemented on a smart hand-held device. This study explores the use of simple bounding box dimensions to predict trunk angle while lifting. Mannequin poses were generated using the Michigan 3DSSPP software for 105 postures across six anthropometries. A regression model for predicting trunk angle was created (adjusted R2=0.91, p<.001). Predicted trunk angles compared against measured 3D motion capture for five participants (N=180 lifts) had a mean error of 15.85° (SE=0.63°, R2 = 0.80). This algorithm should be useful for calculating trunk kinematic properties that are associated with increased risks of low-back disorders including trunk speed and acceleration using successive video frames of predicted trunk flexion angles.

The purpose of this study was to assess two computation models for estimating the hand locations during lifting tasks using data from five inertial measurement units (IMUs) attached to five body segments. The first model computed the hand location with the IMU gyroscope data and the pre-defined ratios of body segment lengths. The second model used the same gyroscope information and all measured lengths of the body segments. The outcome measure of these models was the estimated hand location in 12 lifting zones defined by the ACGIH Threshold Limit Values (TLVs) for lifting. Motion data was collected with the wearable system and a laboratory-grade motion capture system on ten subjects that performed 12 two- handed lifting tasks representing the lifting zones. By including body segment measurements, the average accuracy of the model improved from 4 to 34%, suggesting that body segment information plays an important role in estimating the lifting zones.