Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Machine-learning-based detection of physical activities and body postures using movement sensors
Wearable movement sensors are promising for human activity recognition and the understanding movement behaviour and daily physical activities. In the context of neurologic disease, sensor-derived data provides valuable information to optimize both: prevention strategies and treatment of functional d
ZurichMOVE sensors were developed by the Rehabilitation Engineering Laboratory (RELab, ETH Zurich) and are increasingly used for purposes of rehabilitation research. In previous projects, two machine leaning-based algorithms have been developed, and trained, to detect activities and body postures such as walking, stair ascend/descend, standing, sitting, and lying in a healthy study sample. It remains unclear how accurately these activities/postures are detected in subjects with impairment of gait and balance.
In preparation for this project, a validity study was conducted including 14 stroke patients, who performed a semi-structures protocol of physical activities in their home environment while wearing 5 ZurichMOVE sensors. The ground truth criterion was generated by videotaping and labelling physical activities and body postures, synchronized with movement sensor recordings.
ZurichMOVE sensors were developed by the Rehabilitation Engineering Laboratory (RELab, ETH Zurich) and are increasingly used for purposes of rehabilitation research. In previous projects, two machine leaning-based algorithms have been developed, and trained, to detect activities and body postures such as walking, stair ascend/descend, standing, sitting, and lying in a healthy study sample. It remains unclear how accurately these activities/postures are detected in subjects with impairment of gait and balance.
In preparation for this project, a validity study was conducted including 14 stroke patients, who performed a semi-structures protocol of physical activities in their home environment while wearing 5 ZurichMOVE sensors. The ground truth criterion was generated by videotaping and labelling physical activities and body postures, synchronized with movement sensor recordings.
It is the goal to update and validate two machine learning algorithms to achieve optimal accuracy for gait detection and discrimination between body postures.
It is the goal to update and validate two machine learning algorithms to achieve optimal accuracy for gait detection and discrimination between body postures.
The project includes the following tsks:
• Optimisation/ sequencing of criterion data (labelled data)
• State of the art pre-processing and parameter and feature selection
• Training & validation using support vector machine and K-nearest neighbour -based algorithms
• Evaluation performance and re-adaptation
The project includes the following tsks: • Optimisation/ sequencing of criterion data (labelled data) • State of the art pre-processing and parameter and feature selection • Training & validation using support vector machine and K-nearest neighbour -based algorithms • Evaluation performance and re-adaptation
Requirements
• Experience and sold skills in machine learning approaches
• Advanced skills in programming MATLAB
• Autonomous working and knowledge acquisition
• Of advantage: Interest in the field of sensor-based movement analysis
Requirements • Experience and sold skills in machine learning approaches • Advanced skills in programming MATLAB • Autonomous working and knowledge acquisition • Of advantage: Interest in the field of sensor-based movement analysis
Contact details: Johannes Pohl, cand PhD, MSc. PT Johannes.pohl@usz.ch +41 77 486 8083
Earliest start: October 2021
Institution: University of Zurich and University Hospital Zurich, Department of Neurology,
Collaboration with Rehabilitation Engineering Laboratory, ETH Zurich.
Contact details: Johannes Pohl, cand PhD, MSc. PT Johannes.pohl@usz.ch +41 77 486 8083 Earliest start: October 2021 Institution: University of Zurich and University Hospital Zurich, Department of Neurology, Collaboration with Rehabilitation Engineering Laboratory, ETH Zurich.