Presentation
Multimodal Evaluation of Exoskeleton Assistance Using Optimized Machine Learning
DescriptionRepetitive manual material handling remains a leading occupational risk factor for work-related musculoskeletal disorders, and passive shoulder-support exoskeletons have emerged as promising engineering controls. Most evaluations rely on group-level statistics from single sensor modalities, leaving open whether their effects can be detected at the individual level. The Revised NIOSH Lifting Equation (RNLE) classifies task risk via the Lifting Index (LI), yet few exoskeleton evaluations anchor tasks to RNLE thresholds, and prior machine learning (ML) work has been limited by participant-level aggregation and single-modality inputs. Thirty-five adults performed low-risk (LI < 1.0) and high-risk (LI > 1.0) lifting with and without a passive shoulder-support exoskeleton. Surface EMG, IMU kinematics, and NASA-TLX ratings were extracted into a 34-feature vector and classified using four models under Leave-One-Subject-Out Cross-Validation. SVM (RBF) reached 89.4% accuracy (AUC = 0.932) on low-risk lifting, Logistic Regression reached 84.8% (AUC = 0.904) on high-risk lifting, both exceeding participant-aggregated baselines. NASA-TLX physical demand outranked all biomechanical features. These results show that participant-independent ML on multimodal wearable data can detect exoskeleton effects at the individual level and that perceived workload may be a more consistent cross-subject signal than objective biomechanics, with practical implications for low-burden field evaluation.
Event Type
Lecture
TimeThursday, October 22nd11:40am - 12pm PDT
Location