Technical Details

The center part of the APH-Alarm system is an easy to use mobile applications, suite able for Android Smart Phones (primary application) and an additional secondary application for relatives which will receive incoming emergency calls.

Great effort was put in to the design of both applications, to keep them intuitive and simple. Used icons, especially for the primary application are selected in a way that they are safely recognizable also in stressful situations.

Functionality could be customized through a small and simple subset of configuration switches, to meet the users requirements.

Based on an idea from EvoAid Kft. the consortium the consortium is tuning the AI functionality to provide an emergency APP suitable for people with aphasia after a stroke. Emergency calls are triggered by gesture or by using pictograms, allowing to communicate without the need to speak.

Taking special focus on people burdened by Aphasia this technology has been integrated into a comprehensive safety solution, with the aim to allow triggering alarms in situations where common communication systems fail:

APH-Alarm covers a multitude of interaction possibilities like:

  • Common text interface
  • Symboles
  • Text To Speech
  • Gesture detection
  • Emergency detection 
    • Falls
    • Epileptic seizure

The gesture recognition feature allows to automatically detects specific events like during normal movement data. Trained machine learning models were used to distinguish between actions of daily living and emergency events. The use of this feature requires no additional sensors or other hardware, as all algorithms relay on the build-in sensors of the smart phone. 

To create a reliable system the training data contains thousands of hours of recorded movement date. Data collection is also done during this project, with the help of a special recording application (also part of the projects work, sample image left). Data analysis, annotation and further processing is done at evoAIDs server infrastructure. 

Although the R&D project is already reaching it's end, data collection will be continued to further increase the already great detection rates of the models.

As APH-Alarm aims to be an comprehensive emergency solution, it also tries to cover situations and times where the smartphone is not in reach. Most commonly during night when sleeping. For this case APH-Alarm uses external sensors, wireless connected to the smartphones applications. In comparison to many existing systems, those sensors could be attached to the living area and therefore could not be forgotten to turn on or recharge.

During the R&D project three different ways of sensors have been tested:

  • Indoor Radar
  • Acceleration based sensors (wearable)
  • Acceleration based sensors (fix mounted)

Especially with the indoor radar and the wearable acceleration based sensors more complex gestures could be detected (e.g.: waving, winking, ...). This allows to use APH-Alarm not only to detect emergency situations but to act as a gesture based communication system (see image below).

APH-Alarm in use as gesture based communication system