Capability of Machine Learning Algorithms to Classify Safe and Unsafe Postures during Weight Lifting Tasks Using Inertial Sensors

Diagnostics, Journal Year: 2024, Volume and Issue: 14(6), P. 576 - 576

Published: March 8, 2024

Occupational ergonomics aims to optimize the work environment and enhance both productivity worker well-being. Work-related exposure assessment, such as lifting loads, is a crucial aspect of this discipline, it involves evaluation physical stressors their impact on workers’ health safety, in order prevent development musculoskeletal pathologies. In study, we explore feasibility machine learning (ML) algorithms, fed with time- frequency-domain features extracted from inertial signals (linear acceleration angular velocity), automatically accurately discriminate safe unsafe postures during weight tasks. The were acquired by means one measurement unit (IMU) placed sternums 15 subjects, subsequently segmented extract several features. A supervised dataset, including features, was used feed ML models assess prediction power. Interesting results terms metrics for binary safe/unsafe posture classification obtained logistic regression algorithm, which outperformed others, accuracy area under receiver operating characteristic curve values up 96% 99%, respectively. This result indicates proposed methodology—based single sensor artificial intelligence—to associated load activities. Future investigation wider study population using additional scenarios could confirm potentiality methodology, supporting its applicability occupational field.

Language: Английский

---

Combining Postural Sway Parameters and Machine Learning to Assess Biomechanical Risk Associated with Load-Lifting Activities

Diagnostics, Journal Year: 2025, Volume and Issue: 15(1), P. 105 - 105

Published: Jan. 4, 2025

Background/Objectives: Long-term work-related musculoskeletal disorders are predominantly influenced by factors such as the duration, intensity, and repetitive nature of load lifting. Although traditional ergonomic assessment tools can be effective, they often challenging complex to apply due absence a streamlined, standardized framework. Recently, integrating wearable sensors with artificial intelligence has emerged promising approach effectively monitor mitigate biomechanical risks. This study aimed evaluate potential machine learning models, trained on postural sway metrics derived from an inertial measurement unit (IMU) placed at lumbar region, classify risk levels associated lifting based Revised NIOSH Lifting Equation. Methods: To compute parameters, IMU captured acceleration data in both anteroposterior mediolateral directions, aligning closely body’s center mass. Eight participants undertook two scenarios, each involving twenty consecutive tasks. classifiers were tested utilizing validation strategies, Gradient Boost Tree algorithm achieving highest accuracy Area under ROC Curve 91.2% 94.5%, respectively. Additionally, feature importance analysis was conducted identify most influential parameters directions. Results: The results indicate that combination model offers feasible for predicting risks Conclusions: Further studies broader participant pool varied conditions could enhance applicability this method occupational ergonomics.

Language: Английский

---

Validity of Wearable Inertial Sensors for Gait Analysis: A Systematic Review

Diagnostics, Journal Year: 2024, Volume and Issue: 15(1), P. 36 - 36

Published: Dec. 27, 2024

Background/Objectives: Gait analysis, traditionally performed with lab-based optical motion capture systems, offers high accuracy but is costly and impractical for real-world use. Wearable technologies, especially inertial measurement units (IMUs), enable portable accessible assessments outside the lab, though challenges sensor placement, signal selection, algorithm design can affect accuracy. This systematic review aims to bridge benchmarking gap between IMU-based traditional validating use of wearable systems gait analysis. Methods: examined English studies 2012 2023, retrieved from Scopus database, comparing sensors focusing on IMU body parameters, validation metrics. Exclusion criteria search included conference papers, reviews, unavailable without those not involving agreement or systems. Results: From an initial pool 479 articles, 32 were selected full-text screening. Among them, lower resulted in most common site single placement (in 22 studies), while frequently used multi-sensor configuration involved positioning back, shanks, feet, thighs (10 studies). Regarding 11 out focused spatial-temporal 12 joint kinematics, 2 events, remainder a combination parameters. In terms metrics, 24 employed correlation coefficients as primary measure, 7 error coefficients, Bland–Altman Validation metrics revealed that IMUs exhibited good moderate kinematic measures. contrast, spatiotemporal parameters demonstrated greater variability, ranging poor. Conclusions: highlighted transformative potential advancing analysis beyond constraints laboratory-based

Language: Английский

---

Binary Risk vs No-Risk Classification of Load Lifting Activities Using Features Extracted from sEMG Trapezius Muscle

IFMBE proceedings, Journal Year: 2024, Volume and Issue: unknown, P. 283 - 291

Published: Jan. 1, 2024

Language: Английский

---

Classification Algorithms Trained on Simple (Symmetric) Lifting Data Perform Poorly in Predicting Hand Loads during Complex (Free-Dynamic) Lifting Tasks

Published: Jan. 1, 2024

Language: Английский

---

Enhancing Privacy Protection for Time-Series Signals in Ergonomics Studies via Data Synthesis

Proceedings of the Human Factors and Ergonomics Society Annual Meeting, Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 29, 2024

Electromyography (EMG) signal analysis is critical in understanding and treating work-related musculoskeletal disorders (WMSDs). Despite the increasing use of EMG signals combined with machine learning to assess biomechanical risks various occupational settings, a significant shortage extensive datasets hinders progress. This primarily due stringent data management plans limited availability representing tasks. To address this, our research leverages diffusion models synthesize tailored manual material handling (MMH) tasks, aiming enrich repositories while maintaining privacy. Using conditional model residual U-Net architecture, we synthesized for MMH activities such as pulling, pushing, lifting. The data, evaluated across time frequency domains, demonstrated fidelity original signals, capturing distinct patterns amplitudes characteristic different Our findings highlight potential generating high-fidelity providing novel solution scarcity challenge health research.

Language: Английский

---

Classification algorithms trained on simple (symmetric) lifting data perform poorly in predicting hand loads during complex (free-dynamic) lifting tasks

Applied Ergonomics, Journal Year: 2024, Volume and Issue: 125, P. 104427 - 104427

Published: Dec. 10, 2024

Language: Английский

---

Optimization of Ergonomics in Industrial Companies Based on Artificial Intelligence: Literature Review

Lecture notes in networks and systems, Journal Year: 2024, Volume and Issue: unknown, P. 275 - 296

Published: Jan. 1, 2024

Language: Английский

---