University of Turku

Capstone

Goal

Easy to use web platform for analyzing walking data, which is collected from a smart insole. Users can monitor their walking trends via handy graphs after the data has been uploaded and processed on the server. Additionally we have implemented classifiers for fall detection based on the step data. This can be extended for real time usage. Overall balance for different sessions can also be monitored. 

Implementation

We have selected and generated features from the data which highlight if there are balance problems with the user. We also have trained multiple models for the fall detection task with our data that we collected. The predictions from them are combined with ensemble methdos such as boosting. This implementation can be deployed for use on our Django powered website. 

Future development

- Support for other smart insoles - Extending the platform to mobile for real time monitoring - Possibly calling for help automatically if the user falls
Bild av projektet