University of Twente Student Theses
Evaluation of linear regression models for minimal IMU based estimation of aCOM in gait and non-gait activities
Donck, Mathieu / M van der (2024) Evaluation of linear regression models for minimal IMU based estimation of aCOM in gait and non-gait activities.
|
PDF
16MB |
| Abstract: | An estimation of center of mass acceleration(aCOM) using a small number of inertial measurement units(IMUs) can potentially be used to estimate energy expenditure in daily living for the purpose of monitoring users in their daily lives. IMU data and reference aCOM was gathered for a variety of gait and non-gait activities using the MVN Link suit. Linear regression models were trained on segment acceleration, angular acceleration, angular velocity and orientation data obtained from this IMU data via the MVN record software. The performance of activity specific instantaneous linear regression models for the estimation of aCOM using segment acceleration was investigated for models using inputs from the pelvis segment in addition to one other segment, the pelvis segment in addition to two symmetrical limb segments and the pelvis segment in addition to two symmetrical limb segments and the head or sternum segment. The performance of models using inputs based on the tangential and centripetal acceleration as well as linear acceleration from the pelvis segment in addition to one other segment was also tested. Models using inputs from the head, sternum and shoulder segments performed the best overall. Models using linear acceleration from symmetrical lower leg segments were necessary to obtain better performance than pelvis only models for a balancing task but did not perform well for other activities without the addition of a head or sternum segment. The addition of inputs based on the tangential and centripetal acceleration significantly improved the performance of models using inputs from the upper legs, but such models were still outperformed by models using the head, shoulders or sternum. Large differences in performance for data from different subjects was detected. It was found that such differences were not exacerbated by the addition of inputs based on the tangential and centripetal acceleration. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 44 medicine |
| Programme: | Biomedical Engineering MSc (66226) |
| Link to this item: | https://purl.utwente.nl/essays/103724 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page