Commit e6a228b6 authored by Simon Gölzhäuser's avatar Simon Gölzhäuser
Browse files

Updated proposal README

parent 7aeb5ea3
......@@ -4,22 +4,22 @@ In the file references.bib the references cited in this file in a Latex-like man
## Description
In the literature some work can be found which focuses on detecting hand washing events \cite{kinsella2007} \cite{levchenko2010} \cite{levchenko2011} \cite{fagert2017}. Their overall goal is mostly to improve hand hygiene of healthcare staff in medical environments, like e.g. hospitals.
In the literature some work can be found which focuses on detecting hand washing events \cite{kinsella2007} \cite{levchenko2010} \cite{levchenko2011} \cite{fagert2017}. Their overall goal is mostly to improve hand hygiene of healthcare staff in medical environments, like hospitals.
The hand washing detection and monitoring system which will be built in this master project aims at a different goal. It wants to establish an automatically triggered camera-based ground truth data recording system which can then be used in many different studies. One novel example is the detection of obsessive-compulsive hand washing with wrist-worn devices \cite{scholl2020}.
The collected video data can then in future work also be used to apply gesture recognition Machine Learning approaches to it.
Already existing is a wrist motion data recording software for Android smartwatches \cite{githubSogge} \cite{githubScholl}. Currently there is only limited context in the collected data given by user interaction. No ground truth is automatically recorded. This gap should be filled with:
Already existing is a wrist motion data recording software for Android smartwatches \cite{githubSogge} \cite{githubScholl}. Currently there is no context in the collected data, no ground truth is automatically recorded. This gap should be filled with:
* **Bluetooth beacons** placed at sinks, which communicate with the smartwatches. Start and endpoints of hand washing activities (as ground truth) can then be detected with rough accuracy instead of asking the user for it. This gives primary evidence.
* **Bluetooth beacons** placed at sinks, which communicate with the smartwatches. Start and endpoints of hand washing activities (as ground truth) can then be detected with rough accuracy instead of asking the user for it. This gives a primary evidence.
* An **automatically triggered video recording system integrated into an electric soap dispenser**. After placing this device appropriately at a sink it should record ground truth video data each time when someone is washing hands. Timestamp information can then be used to associate the videos with the wrist motion data from the smartwatch. This provides a secondary evidence.
* An **automatically triggered video recording system integrated into an electric soap dispenser**. After placing this device appropriately at a sink it should record ground truth video data each time when someone is washing hands. Timestamp information can then be used to associate the videos with the wrist motion data from the smartwatch. This provides secondary evidence.
* A **flow rate sensor** added to the soap dispenser setup. This should then be utilized to add more accuracy to the detection of hand washing events
These are also the three levels of complexity the work will be divided in. The final setup could then be used for studies addressing all kinds of hand washing activities.
The collected video data can then in future work also be used to create ground truth data for any kind of gesture recognition Machine Learning approaches based on the wrist motion data. Or also to apply Machine Learning to the video data itself.
## Roadmap & Deliverables
- [ ] **May:** Integration of Bluetooth beacons in wrist motion recording software
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment