Author Contributions
Conceptualization, P.G. and A.F.; methodology, P.G., G.B., A.F.; software, P.G., G.B., C.A.A.; validation, P.G., G.B. and A.F.; investigation, P.G., G.B., and A.F.; data curation, P.G. and G.B.; writing—original draft preparation, P.G., G.B., and A.F.; writing—review and editing, G.B. and A.F.; supervision, C.A.A. and A.F.; project administration, A.F.; funding acquisition, A.F. All authors have read and agreed to the published version of the manuscript.
Figure 1.
System Architecture. On the left side wearable devices and the communication with a host PC. On the right side the modules that compose the system, from the task analysis to the online segmentation and processing of kinematic and EMG data to obtain the risk scores. Green blocks represent software modules where the input of the rater is required. Yellow blocks represent modules that provide scores and output of the analyses available to the rater.
Figure 1.
System Architecture. On the left side wearable devices and the communication with a host PC. On the right side the modules that compose the system, from the task analysis to the online segmentation and processing of kinematic and EMG data to obtain the risk scores. Green blocks represent software modules where the input of the rater is required. Yellow blocks represent modules that provide scores and output of the analyses available to the rater.
Figure 2.
Flowchart of the risk evaluation procedure for a new working activity. The procedure takes into account possible iterations on the training of the neural network for automatic segmentation of the activity.
Figure 2.
Flowchart of the risk evaluation procedure for a new working activity. The procedure takes into account possible iterations on the training of the neural network for automatic segmentation of the activity.
Figure 3.
Wearable inertial devices worn by volunteers.
Figure 3.
Wearable inertial devices worn by volunteers.
Figure 5.
Kinematic chain used for motion reconstruction. Centres of the spherical joints coincide with the origins of the frame attached to the child link following a hierarchical structure rooted in the waist. The names of these origins for some relevant frames are reported. Only the right side of the human body has been reported for symmetric links. The listed joints refer to: root (Root), chest (Che), neck (Ne), head (He), collar (Cl), shoulder (Sh), elbow (El), wrist (Wr), hip (Hp), knee (Kn), ankle (An) and toe (Toe)
Figure 5.
Kinematic chain used for motion reconstruction. Centres of the spherical joints coincide with the origins of the frame attached to the child link following a hierarchical structure rooted in the waist. The names of these origins for some relevant frames are reported. Only the right side of the human body has been reported for symmetric links. The listed joints refer to: root (Root), chest (Che), neck (Ne), head (He), collar (Cl), shoulder (Sh), elbow (El), wrist (Wr), hip (Hp), knee (Kn), ankle (An) and toe (Toe)
Figure 6.
Comparison between the devices latencies. Listed below there are the acronyms used: Analog to Digital Converter (ADC), Serial Peripheral Interface (SPI), Universal Asynchronous Receiver Transmitter (UART), Bluetooth (BLT) and Moven Software (MVN SW).
Figure 6.
Comparison between the devices latencies. Listed below there are the acronyms used: Analog to Digital Converter (ADC), Serial Peripheral Interface (SPI), Universal Asynchronous Receiver Transmitter (UART), Bluetooth (BLT) and Moven Software (MVN SW).
Figure 7.
NIOSH Interface.
Figure 7.
NIOSH Interface.
Figure 8.
Foreseen temporal structure and duration comparison of risk evaluation for the Traditional and new method. Column groups report traditional evaluation on the left column and the proposed method on the right column. The assumption that an expert rater does not need to annotate the whole recorded activity to score the risk has been adopted.
Figure 8.
Foreseen temporal structure and duration comparison of risk evaluation for the Traditional and new method. Column groups report traditional evaluation on the left column and the proposed method on the right column. The assumption that an expert rater does not need to annotate the whole recorded activity to score the risk has been adopted.
Figure 9.
Lift-on/Lift-off (LoLo) phases and ergonomic methods selected.
Figure 9.
Lift-on/Lift-off (LoLo) phases and ergonomic methods selected.
Figure 10.
A volunteer ready to start his work shift wearing the wearable devices. The sensor units have been placed on the body segments as described in
Section 3.4.1.
Figure 10.
A volunteer ready to start his work shift wearing the wearable devices. The sensor units have been placed on the body segments as described in
Section 3.4.1.
Figure 11.
Multi-Layer Perceptron (MLP) Neural Network performance. The rows refer to the predicted class (Output Class) and the columns correspond to the true class (Target Class). The last column represents the percentages of all inputs correctly and incorrectly predicted for each class in green and red respectively. The last row represents the percentages relative to the target class that are correctly and incorrectly classified in green and red respectively. In the inner 3×3 square, the diagonal cells green squares correspond to correctly classified inputs, the other cells (red squares), to incorrecly classified inputs. The black numbers and percentages refer to the number of elements for each class and their percentage over the total dataset for the network test. The overall accuracy is reported in the lowest and rightest cell.
Figure 11.
Multi-Layer Perceptron (MLP) Neural Network performance. The rows refer to the predicted class (Output Class) and the columns correspond to the true class (Target Class). The last column represents the percentages of all inputs correctly and incorrectly predicted for each class in green and red respectively. The last row represents the percentages relative to the target class that are correctly and incorrectly classified in green and red respectively. In the inner 3×3 square, the diagonal cells green squares correspond to correctly classified inputs, the other cells (red squares), to incorrecly classified inputs. The black numbers and percentages refer to the number of elements for each class and their percentage over the total dataset for the network test. The overall accuracy is reported in the lowest and rightest cell.
Figure 12.
NIOSH scores for lasching and unlashing in phase 1 and phase 2. The black columns represent the average and the standard deviation of the NIOSH score with the proposed method. The average scores and their standard deviations obtained through the manual evaluation are represented in red. Three horizontal dashed lines represent the different risk thresholds: a possible risk in yellow, a high risk in orange and very high risk in red.
Figure 12.
NIOSH scores for lasching and unlashing in phase 1 and phase 2. The black columns represent the average and the standard deviation of the NIOSH score with the proposed method. The average scores and their standard deviations obtained through the manual evaluation are represented in red. Three horizontal dashed lines represent the different risk thresholds: a possible risk in yellow, a high risk in orange and very high risk in red.
Figure 13.
Factors that determines the Lifting Index in the two phases of Lashing and Unlashing operations. The background colors indicate the average factor values that, under the assumption , would produce a LI score corresponding to low (green), high (yellow t orange) and very high (red) risk.
Figure 13.
Factors that determines the Lifting Index in the two phases of Lashing and Unlashing operations. The background colors indicate the average factor values that, under the assumption , would produce a LI score corresponding to low (green), high (yellow t orange) and very high (red) risk.
Figure 14.
Distribution of scores for the postural factors of the REBA score. The background color varies between green (minimum value of the score) and red (maximum value of the score).
Figure 14.
Distribution of scores for the postural factors of the REBA score. The background color varies between green (minimum value of the score) and red (maximum value of the score).
Figure 15.
REBA score for lashing and unlashing in phase 3. The black columns represent the average and the standard deviation of the REBA score with the proposed method. The average scores and their standard deviations obtained through the manual evaluation are represented in red. The three horizontal dashed lines represent the different risk thresholds: a medium risk in yellow, a high risk in orange and very high risk in red.
Figure 15.
REBA score for lashing and unlashing in phase 3. The black columns represent the average and the standard deviation of the REBA score with the proposed method. The average scores and their standard deviations obtained through the manual evaluation are represented in red. The three horizontal dashed lines represent the different risk thresholds: a medium risk in yellow, a high risk in orange and very high risk in red.
Figure 16.
Strain Index score for lashing and unlashing in phase 3. The black columns represent the average and the standard deviation of the SI score with the proposed method. Three horizontal dashed lines represent the different risk thresholds: a low risk in yellow, a high risk in orange and very high risk in red.
Figure 16.
Strain Index score for lashing and unlashing in phase 3. The black columns represent the average and the standard deviation of the SI score with the proposed method. Three horizontal dashed lines represent the different risk thresholds: a low risk in yellow, a high risk in orange and very high risk in red.
Table 1.
Characteristics of the three inertial systems presented: Xsens, Perception Neuron, and custom device.
Table 1.
Characteristics of the three inertial systems presented: Xsens, Perception Neuron, and custom device.
Characteristics | Xsens | Perception Neuron | Custom |
---|
Suggested amount of IMUs | 17 | 17 | 11 |
Transmission frequency [Hz] | 240 | 120 | 100 |
Wireless protocol | Wi-Fi | Wi-Fi | Wi-Fi |
Latency [ms] | 20 | 10–13 | 3 |
Battery life [h] | 9.5 | 3.5 | 5 |
Buffering [min] | 10 | - | - |
Weight [kg] | 1.13 | 0.85 | 0.54 |
Magnetic disturbances tolerant | yes | no | no |
Table 2.
Characteristics of the two EMG systems presented: Shimmer3 and custom.
Table 2.
Characteristics of the two EMG systems presented: Shimmer3 and custom.
Characteristics | Shimmer3 | Custom |
---|
Suggested units number | 2 | 2 |
Differential channels | 4 | 4 |
Wireless protocol | Bluetooth | Wi-Fi |
Transmission frequency [kHz] | 0.125–8 | 0.250–32 |
Resolution [bit] | 12 | 24 |
CMRR [dB] | - | −115 |
SNR [dB] | - | 15–16 dB |
SD card [GB] | 8 | - |
Gain | 1–12 | 1–12 |
Dimension [cm] | 6.5 × 3.2 × 1.2 | 6 × 5 × 1.5 & 4 × 5 × 0.5 |
Latency [ms] | 25 | 3 |
Battery life [h] | 24 | 10 |
Unit weight [g] | 45 | 110 |
Table 3.
The NIOSH table represents the parameters that must be set manually and those calculated. Abbreviations: Y stands for YES, N stands for NO.
Table 3.
The NIOSH table represents the parameters that must be set manually and those calculated. Abbreviations: Y stands for YES, N stands for NO.
NIOSH |
---|
| Task Analysis Input | Calculated |
---|
Age and Gender | Y | N |
Lifted Mass | Y | N |
Horz. Distance | N | Y (motion capture) |
Vertical Location | N | Y (motion capture) |
Vertical Displacement | N | Y (motion capture) |
Asymmetry Angle | N | Y (motion capture) |
Frequency | N | Y(motion capture) |
Quality of the Grip | Y | N |
Table 4.
The Snook & Ciriello table represents the parameters that must be set manually and those calculated. Abbreviations: Y stands for YES, N stands for NO.
Table 4.
The Snook & Ciriello table represents the parameters that must be set manually and those calculated. Abbreviations: Y stands for YES, N stands for NO.
Snook & Ciriello |
---|
| Task Analysis Input | Calculated |
---|
Gender | Y | N |
The sustained and initial force | Y | N |
Handle height | N | Y (motion capture) |
Covered Distance | N | Y (motion capture) |
Frequency | N | Y(motion capture) |
Table 5.
The REBA table represents the steps for REBA score evaluation and the parameters that must be set manually and those calculated. Abbreviations: Y stands for YES, N stands for NO.
Table 5.
The REBA table represents the steps for REBA score evaluation and the parameters that must be set manually and those calculated. Abbreviations: Y stands for YES, N stands for NO.
REBA |
---|
Step | Description | Variable | Task Analysis Input | Calculated |
---|
1 | Neck Position | Head Flexion | N | Y (motion capture) |
Head is twisted | N | Y (motion capture) |
Head is side bending | N | Y (motion capture) |
2 | Trunk Position | Trunk Flexion | N | Y (motion capture) |
Trunk is twisted | N | Y (motion capture) |
Trunk is side bending | N | Y (motion capture) |
3 | Legs | Legs Flexion | N | Y (motion capture) |
4 | Posture Score A | Table A | N | Y (motion capture) |
5 | Load Score | Load Class | Y | N |
Shock or rapid build up of force | Y | N |
6 | Score A | Posture Score A + Load Score | N | Y (from table) |
7 | Upper Arm | Shoulder Flexion | N | Y (motion capture) |
Shoulder is raised | N | Y (motion capture) |
Upper Arm is abducted | N | Y (motion capture) |
Person is leaning | N | Y (motion capture) |
8 | Lower Arm Position | Elbow Flexion | N | Y (motion capture) |
9 | Wrist Position | Wrist Flexion | N | Y (motion capture) |
Wrist is bent | N | Y (motion capture) |
Wrist is twisted | N | Y (motion capture) |
10 | Posture Score B | Table B | N | Y (from table) |
11 | Add Coupling Score | Quality of the Grip | Y | N |
12 | Score C | Table C | N | Y (from table) |
13 | ctivity Score | Static position (longer than 1 min) | N | Y (motion capture ) |
Repeated Action (more 4× per minute) | N | Y (motion capture) |
Action causes rapid large range changes | Y | N |
Output | Final Score | Score C + Activity Score | N | Y (from table) |
Table 6.
The Strain Index table represents the parameters that must be set manually and those calculated.Abbreviations: Y stands for YES, N stands for NO.
Table 6.
The Strain Index table represents the parameters that must be set manually and those calculated.Abbreviations: Y stands for YES, N stands for NO.
Strain Index |
---|
| Task Analysis Input | Calculated |
---|
Intensity of Exertion | N | Y (EMG) |
Duration of Exertion | N | Y (EMG) |
Exertions per minute | N | Y (EMG) |
Wrist Posture | Y (for grip quality) | Y (motion capture for wrist posture) |
Speed of Work | Y | N |
Duration per Day | Y | N |