Seminar/Thesis: machine learning algorithms for context recognition using mobile sensors
Wearables, including smartwatches, as well as smartphones are equipped with several sensors, which allow monitoring of data relevant to retrieve digital biomarkers. Some of the biomarkers measured by wearables are already common in daily life (e.g. step counters), which foreshadows a future with even more extensive self-quantification in the health sector. To evaluate current possibilities of commercial devices and their sensors, particularly in the field of dietary monitoring, this seminar aims at developing a machine learning pipeline for the development of context recognition systems, which detect daily life situations relevant for dietary assessment. The algorithms will be implemented using pre-existing datasets, and shall be deployed in a smartphone app using e.g. Tensorflow Lite.
Understand the potential of mobile sensors; develop ML algorithm using pre-existing datasets
- Understand sensor types in smartphones and smartwatches for digital health
- Develop machine learning algorithm for context recognition
- Deploy algorithm on a smartphone app
|ECTS||2.5, 5, 7.5|
|Language||English and/or German|
|Period||Winter term 2021/22|
|Presence time||Virtual seminar, working from remote|
|Useful knowledge||Machine learning, app development, Python, datasets|
|Work distribution||20% data investigation and literature research 80% programming in Python|
|Med. Eng. designation||Advanced Context Recognition (ACR)|
|StudOn link||Link will follow shortly.|
|First meeting||Online introduction/Vorbesprechung|
|Registration||Via StudOn, obligatory after introduction.|
Up-to-date literature recommendations are provided during the meetings.
Final presentation and final report.