Loading...

Measurements of the Upper Trunk and Pelvis kinematics Due to Two-handed Symmetric and Asymmetric Reach and Lifting Activities Using Inertial Sensors and Presenting a Neural Network for Posture Prediction

Gholipour, Alireza | 2015

572 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 47843 (08)
  4. University: Sharif University of Technology
  5. Department: Mechanical Engineering
  6. Advisor(s): Arjmand, Navid; Parnianpour, Mohammad
  7. Abstract:
  8. Manual material handling (MMH) activities are identified as risk factors for occupational low back pain (LBP). Task-related variables including load and posture characteristics are required as input into these models and tools for estimation of trunk external moment. Biomechanical modeling studies that aim to mathematically estimate low back loads currently need an inevitable parallel time-consuming in vivo study in an equipped laboratory to measure trunk posture under physical activities. Inertial sensors as a portable, accurate, almost inexpensive, and small device could be very helpful in order to capture the kinematic data of movement in human activities. My master thesis aims to investigate capability of the ANNs in predicting spinal posture (defined by three sets of 3D angles, i.e., pelvis or S1, T12 and upper trunk or T1 orientations in the space) during various symmetric and asymmetric reach and load handling activities for a male population having smaller, compared to the previous studies, inter-individual variations in terms of age, body height and body mass index. Two distinct ANNs to predict reach and lifting activities are trained and tested to predict 3D orientation of the pelvis, lumbar, and thorax based on the in vivo measurements using an inertial tracking device (Xsens MTx, Xsens Technologies, Enschede, Netherlands). Inputs of each ANN is the 3D position of hand load with subject height while its outputs are nine Eulerian angles of the segments under consideration (i.e., S1, T12, and T1). It is hypothesized that such population-based ANNs, rather than subject-specific ones, are robust tools to accurately predict spinal posture for a given hand load position during 3D reach and load handling activities. These ANNs could be easily used to predict spinal posture for use in musculoskeletal biomechanical models of the spine to estimate risk of injury based on the predicted spinal loads and muscle forces
  9. Keywords:
  10. Artificial Neural Network ; Posture Prediction ; Spine ; Kinematics ; Static Lifting ; Inertial Tracking Sensors

 Digital Object List

 Bookmark

No TOC