Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Master Thesis: Event Segmentation and Detection in Time-Series for Monitoring Activities of Daily Living in SCI Individuals
This thesis explores precise event segmentation in time-series/video data from wearable sensors to monitor daily activities in spinal cord injury individuals.
**Background**
Spinal cord injuries (SCI) pose significant challenges, impacting millions globally and often leading to reduced mobility and difficulties in daily activities. Monitoring activities of daily living (ADLs) is crucial for assessing overall health. Wearable sensors like IMU sensors, pressure mats, and cameras provide wealth of data showing promise for monitoring daily life, however, intelligently segmenting daily behaviors and activities remains a hurdle. This thesis focuses on researching event segmentation and detection techniques on time-series data from diverse signal modalities and video data. Precise segmentation aims to reveal patterns, anomalies, and critical events, serving as key indicators for individuals' well-being. Accurate event detection holds potential for personalized healthcare, enabling timely interventions and tailored care plans.
**Useful References**
Find useful resources in the attached .pdf file.
**Your Tasks**
1. Dive into existing research on activity recognition, with a focus on event detection and data segmentation.
2. Construct a robust pipeline using sliding ADL classification models. After processing the measurements, fine-tune the model output for precise ADL segmentation.
3. Implement state-of-the-art event detection and data segmentation methods for time-series data such as evolutionary state graph [1]. Compare these advanced techniques with the traditional sliding approach for ADL monitoring.
4. Evaluate the outputs of the ADL segmentation pipeline for outpatient monitoring, focusing on identifying and establishing benchmarks for unknown classes in the data.
5. Craft a concise report summarizing the project’s findings.
6. Preparation of a manuscript showcasing the project's findings, for journal submission (optional).
**Your Benefits**
Gain unique access and first-hand experience in one of the leading institutions on long-term health management - At the Swiss Paraplegic Center at Nottwil and ETH Zurich.
**Your Profile**
- Strong experience with Python
- Understanding of multi-modal machine learning methods.
- Knowledge of virtual environments (conda / docker)
- Structured and reliable working style
- Ability to work independently on a challenging topic
**Background**
Spinal cord injuries (SCI) pose significant challenges, impacting millions globally and often leading to reduced mobility and difficulties in daily activities. Monitoring activities of daily living (ADLs) is crucial for assessing overall health. Wearable sensors like IMU sensors, pressure mats, and cameras provide wealth of data showing promise for monitoring daily life, however, intelligently segmenting daily behaviors and activities remains a hurdle. This thesis focuses on researching event segmentation and detection techniques on time-series data from diverse signal modalities and video data. Precise segmentation aims to reveal patterns, anomalies, and critical events, serving as key indicators for individuals' well-being. Accurate event detection holds potential for personalized healthcare, enabling timely interventions and tailored care plans.
**Useful References**
Find useful resources in the attached .pdf file.
**Your Tasks**
1. Dive into existing research on activity recognition, with a focus on event detection and data segmentation. 2. Construct a robust pipeline using sliding ADL classification models. After processing the measurements, fine-tune the model output for precise ADL segmentation. 3. Implement state-of-the-art event detection and data segmentation methods for time-series data such as evolutionary state graph [1]. Compare these advanced techniques with the traditional sliding approach for ADL monitoring. 4. Evaluate the outputs of the ADL segmentation pipeline for outpatient monitoring, focusing on identifying and establishing benchmarks for unknown classes in the data. 5. Craft a concise report summarizing the project’s findings. 6. Preparation of a manuscript showcasing the project's findings, for journal submission (optional).
**Your Benefits**
Gain unique access and first-hand experience in one of the leading institutions on long-term health management - At the Swiss Paraplegic Center at Nottwil and ETH Zurich.
**Your Profile** - Strong experience with Python - Understanding of multi-modal machine learning methods. - Knowledge of virtual environments (conda / docker) - Structured and reliable working style - Ability to work independently on a challenging topic
The primary objective of this project is to accurately identify and classify distinct activity segments and the specific types of activities within those segments. The study will utilize 48 hours of daily life data from 15 individuals with spinal cord injuries, collected through accelerometers, gyroscopes, pressure mats, and video recordings.
The primary objective of this project is to accurately identify and classify distinct activity segments and the specific types of activities within those segments. The study will utilize 48 hours of daily life data from 15 individuals with spinal cord injuries, collected through accelerometers, gyroscopes, pressure mats, and video recordings.
Hosts: Mehdi Ejtehadi, Dr. Diego Paez (SCAI-Lab, ETHZ | SPZ)
Please send your CV and the latest transcript of records from my studies to Mehdi Ejtehadi and Dr. Diego Paez (mehdi.ejtehadi@hest.ethz.ch, diego.paez@hest.ethz.ch)
Hosts: Mehdi Ejtehadi, Dr. Diego Paez (SCAI-Lab, ETHZ | SPZ)
Please send your CV and the latest transcript of records from my studies to Mehdi Ejtehadi and Dr. Diego Paez (mehdi.ejtehadi@hest.ethz.ch, diego.paez@hest.ethz.ch)
adl-segmentation.pdf
