Machine learning and IMU: Smart Gait applications for use in real-time systems
Master thesis
Permanent lenke
https://hdl.handle.net/11250/3129809Utgivelsesdato
2024Metadata
Vis full innførselSamlinger
Sammendrag
Inertial Measurement Units hold a great potential for real-time gait phase measurements, gaitkinematics measurement, and gait analysis. They offer convenience in the form of their light weightand portability. There are however known drawbacks in the form of sensor drift and susceptibilityto noise.Extrapolating data for gait analysis directly from IMU’s can therefore be problematic, but emergingMachine learning / Artificial intelligence (ML/AI) modalities holds the potential of estimatinggait characteristics to a new level of accuracy. Applications such as interpolation of user intent inprosthetic device control could through this be improved further, as well as movement analysis, andgait identification (GI) applications.In this thesis, IMU sensor data is used in combination with machine learning (ML) models toextrapolate significant information about the users walking gait. The information is extrapolatedimplicitly, through Machine Learning algorithms. The is centered around two applications: Gaitdetection (GD, i.e Gait phase detection), and Gait Identification (GI, i.e identify human subjectsbased on their walking gait). Both applications is already present in the scientific literature. Thecomputational load for some of the systems is however quite high, and not fit for a live systemrunning on a microcontroller or a mobile device. The proposed solution attempts to eliminatesome of the computational load by reducing the dimension of input data, and reducing ML modelcomplexity.The work yields two implementations of Machine Learning on a Gait Identity task, and includesdiscussion and analysis of important features for a ML Gait detection task for a dataset consistingof 18 features.For the first Gait Identity task, a TCN ML model architecture scored a perfect 1.0 on a MaxF1-Averaged Stratified k-Fold validation for k=5 on a self-acquired single-IMU dataset with 13subjects.On the second Gait Identity task, a TCN ML model architecture scored a 0.98 on a Max F1-AveragedStratified k-Fold validation for k=5 on a publicly available single-IMU dataset with 30 subject