Neuromuscular-kinematics machine learning models of nonlinear locomotion initiation
This study presents a systematic research on machine learning of neuromuscular-kinematics data for advanced prediction of nonlinear locomotion. This addresses an emergent issue in high-performance non-linear human-robotics interactions with gait training or assistive exoskeletons. To this end, we fi...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Master by Coursework |
| Language: | English |
| Published: |
Nanyang Technological University
2025
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/182359 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | This study presents a systematic research on machine learning of neuromuscular-kinematics data for advanced prediction of nonlinear locomotion. This addresses an emergent issue in high-performance non-linear human-robotics interactions with gait training or assistive exoskeletons. To this end, we first examine human motion data involving multiple-channel electromyogram signal and inertial measurements. Especially, we propose an algorithm to automatically determine the optimal Predictive Lead Times (PLTs). We demonstrate that the PLTs differ significantly between left vs right turning motions on the same stance foot. We conduct Bayesian-based analysis to examine the statistical significance of distinguishable sEMG prior to the motion onset. Subsequently, we introduce long-short-term-memory to the recursive processing and prediction of the continuous neuromuscular process starting from the idle state. Finally, we examine the relationship between data quality and machine learning performance. We demonstrate that, by rejecting corrupted trials by e.g. motion within the designated idle state, the prediction performance can be considerably improved. |
|---|