Open Collections

Zhang, Wenwen

University of British Columbia

The pervasiveness of wearable sensors has contributed to plenty of daily activity data and greatly improved the ability to link data analytics with healthcare applications. Gait analysis is an important part of the healthcare field to help monitor the patient’s progress with gait disturbances such as Stroke and Parkinson’s. The traditional method to get gait parameters is the GaitRite walkway, which is cumbersome and requires professional training in setting up every time. With the assistance of wearable sensors, we are able to record the motion of the foot without those limitations. Zero Update Position and Timing (ZUPT) has been used to analysis of normal people’s walking for a long time. However, ZUPT has been validated on the walking of normal people only. In this article, we test the ZUPT on pathological patients. We compare the results of normal and abnormal walking patterns in terms of step recognition and gait parameter prediction accuracy. To help track and monitor patient disease progress with better accuracy, we propose Machine Learning (ML) algorithms to improve the accuracy of healthcare application estimation results based on the time series data. Those patients also suffer from gait impairments which may cause fatal dangers such as falls. In the meantime, gait parameters are also significant signs of the progress of the disease. To monitor the gait condition, we collect datasets with different gaits, comparing and predicting gait parameters for the patients. Compared to the ZUPT algorithm, the final prediction from ML algorithm of gait parameters achieves greater performance.

Text

2023-05-03

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/84558

Electrical and Computer Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/