Subject

Context: Performing gait analysis and understanding/computing gait spatiotemporal parameters (SGPs) is crucial as it provides valuable insights into an individual motion, balance and overall health status. Gait events, such as initial contact (IC), final contact (FC), side (left/right) detection, are key in computing SGPs from IMUs. Challenges in accurate gait event detection arise from variability in human movements, signal noise, sensor placement, no proper ground truth validation etc, making the development of robust algorithms essential for effective rehabilitation and monitoring of disease progression.

Kind of work

Objective: To develop and evaluate DL models (particularly transformer network/s) for gait event detection using IMU data. This includes developing a model to detect ICs, FCs and left and right sides.

Description of Work:

Literature review (ETOC: 2 months): Literature review of existing DL based methods for gait event detections of sacrum based IMU data.

Dataset Familiarization (ETOC: 1 month): Understand the IMU datasets to be used, especially focusing on this dataset from Zenodo, which includes validated ground truth for gait events. Analyse the dataset’s structure, signal quality, and suitability for training deep learning models.

Algorithm implementation (ETOC: 6 months): Develop transformer networks for detecting gait events. The initial model will be trained using available open-source datasets and then fine tune with the specified dataset to provide gait event detections.

Framework of the Thesis

Related work:

[1] Prakash, C., Kumar, R., & Mittal, N. (2018). Recent developments in human gait research: parameters, approaches, applications, machine learning techniques, datasets and challenges. Artificial Intelligence Review, 49, 1-40.

[2] Tan, T., Shull, P. B., Hicks, J. L., Uhlrich, S. D., & Chaudhari, A. S. (2024). Self-Supervised Learning Improves Accuracy and Data Efficiency for IMU-Based Ground Reaction Force Estimation. IEEE Transactions on Biomedical Engineering.

[3] Küderle, A., Roth, N., Zlatanovic, J., Zrenner, M., Eskofier, B., & Kluge, F. (2022). The placement of foot-mounted IMU sensors does affect the accuracy of spatial parameters during regular walking. Plos one, 17(6), e0269567.

[4] Palmerini, Luca, et al. "Mobility recorded by wearable devices and gold standards: the Mobilise-D procedure for data standardization." Scientific Data 10.1 (2023): 38.

[5] Alharthi, A. S., Yunas, S. U., & Ozanyan, K. B. (2019). Deep learning for monitoring of human gait: A review. IEEE Sensors Journal, 19(21), 9575-9591.

Number of Students

1

Expected Student Profile

Following an MSc in a field related to one or more of the following: Computer Science, Biomedical Engineering, Applied Computer Science - Digital Health.

Strong programming skills (Python).

Experience with signal processing and DL.

Interest/Motivation in developing state-of-the-art DL methods and conduct experiments.

Ability to write scientific reports and communicate research results at conferences in English.