It is known to medical science that falls leading to grave or even fatal injury are a great challenge that cannot be neglected, especially for elderly people. In our project, we created a smartphone-app which detects when someone falls. The app must be installed and activated, then once it's up and running it continues to monitor. If the app/AI detects a possible fall by unusual acceleration or rotation of the mobile, it opens itself on the surface screen. In case of an inaccurate detection of a fall, a "false alarm" so to speak, the person carrying the mobile can decline the accuracy of fall on the touchscreen. In case that a person has fallen but needs no help the person can also click on the screen function: “Deactivate progress” to inactivate the time counter which will at the end of the countdown automatically send an emergency call. In case of a false fall detection where the counter is stopped by the user an additional feedback is given to the AI program. In this way the AI keeps learning and becomes more precise. If the system receives no "answer" by the mobile carrier, it is very likely that a fall has occurred and the person is in need of help due to injury and/or unconsciousness. In this case, after a certain time frame (30 seconds) the app sends an emergency call, by itself, to an emergency hotline with the actual coordinates and data of the phone.
*Source: https://pixabay.com/de/vectors/a-ich-ai-anatomie-2729794/ Date: 27.09.19
| More poject details | Code of the model | Code of the evaluation program | The app | The dataset | Accuracy | The team | Our Video |