A revolutionary smartphone safety app that harnesses Artificial Intelligence (AI) to recognise distress in its user’s activity has been created by a team at University of Bath.
The Epowar app, trials of which started this week, monitors a wearer’s heart rate and body motion.
When changes in these are detected it automatically sends an emergency alert to the smartphone’s contactsand sets off an alarm.
Aimed primarily at women’s safety, Epowar gets around the key problem with rape alarms that need to be physically activated – something often not possible in the event of an attack.
Epowar solves this problem by using AI. In the event of an attack, it senses distress, sends an alert to the wearer’s contacts, sets off a loud alarm as well as recording evidence that will be stored immediately in a cloud system.
E-J Roodt, a BSc Business student in the university’s School of Management , came up with the idea while jogging in a badly lit park and worrying about the risk of an attack.
A keen smartwatch user, she was aware of the advancements in wearable technology and how it was being used to detect heart attacks – saving real lives.
She wondered if those concepts could be applied to women’s safety. After taking her ideas to Maks Rahman, an engineering student who had just returned from a year at medical research organisation Fraunhofer IPA, they co-founded Epowar.
Roodt said: “After months of research and experiment, we were fascinated to find that people’s responses to distress were remarkably consistent and that this could be reliably captured and interpreted using AI.
“We have now reached the development stage where in-field testing can start and are confident we are close to creating a finished product.”
The AI-powered system was built on extensive research into detectable responses to physical distress and an analysis of thousands of samples of physiological and motion data. The AI models can distinguish between physical and psychological stress.
Roodt added: “It occurred to us that a smartwatch with this app may be a way to alert others if a woman is restrained or struggling.
“The key is that it would all happen automatically and an assailant would have little or no time to prevent this – which is not always possible with conventional panic buttons, rape alarms or your mobile phone.”
Roodt is keen to avoid privacy issues that have dogged other security apps which may include tracking the user.
Epowar’s software does not track or identify the wearer, up to the point where an alert is issued. A user could choose to run the app permanently or switch it on for specific times when they felt there was potential risk.
The team behind the app have assembled a number of volunteers who, among other exercises, will simulate physical attack to fine-tune the software. The data collected would be used to fine-tune the app but would be anonymised, she said.
Roodt said: “We are keen to find ways to make this as affordable and accessible to as many women as possible and could envisage a system where organisations, such as schools or universities, make available such software to groups for example.
“We hope people will recognise the ability to automatically alert contacts as a game-changer in a world where such software seems increasingly necessary.”
