AI comes to your Rescue – Smart Safety Products minus Privacy Issues

These days, the AI (Artifical Intelligence) label is being stuck to everything you interact with from digital assistants on your phones and speakers, self-driving cars, social media which automatically recognise and tag your photos, and yes even AI-powered toaster ovens.  One application which caught my attention recently was the application of AI in safety and security.

The work which Sanjiv Bhakat and the team at Wise Systech are doing in the field looks promising. Their product uses AI to scan the surroundings for signs of violence and triggers for help automatically. Sanjiv and team have been working for the past few years towards developing a solution to curb violence against women. Their product called IRA ( Intelligent Response Application) is Mr.Bhakat’s crusade to contribute towards creating safer spaces through pre-emption and instant reaction.

To get a quick run-through of IRA, I had a quick Skype call with  Arik Faingold, President,  and Nogah Hender, VP Product. They took me through how IRA works and gave some fascinating insights on how they harness the power of Artificial Intelligence to improve safety & security.  They are currently at demoing IRA at CES, and there are links to the videos of their CES demos at the end of this article.

Proactive Detection for Autonomous security

Most security measures like CCTV feeds are only useful after an event has occurred to look for proof that it happened. Even with a well spread out network of cameras scanning the surroundings, unless someone is looking at these screens at all times, you can’t take action about theft or violent act while it happens.

Dangerous Situation

IRA shows promises in this field since both their products IRA Voice and IRA Visual uses the results of Computer Vision, Multimedia Fusion & Deep Learning to learn how to identify intrusions, dangerous objects and violent behaviour. I’ll explain some of these keywords below. Since the product detects these signs without any human interaction, the system can notify users and call for help early.

They will be piloting it first in the Asia Pacific region, which includes India where it’ll be rolled out in the first phase in May 2018.

IRA Voice

IRA Voice is a Personal Safety and Emergency app targetted primarily as a Women’s Safety which runs on a regular smartphone.

During the setup process, the actions to trigger in case of emergencies are configured. This could be alerting friends and family or even the police.  The app when activated and kept in the background keeps listening to the surrounding voices and sounds for signs of danger. Since the AI component of the app analyses the audio from the surroundings when it’s launched, if it detects any signs of violence or emergencies, the distressed user needn’t take any overt action for it to trigger the pre-configured responses.

Since the audio processing is done within the app, the audio is not sent anywhere outside the phone. This is especially useful since the app is not dependant on an internet connection being active and allows for quick action in times of danger.  This also avoids invading the privacy of the user during normal use.

IRA Visual

IRA Visual consists of a smart camera which can be deployed at homes or offices. This device uses  Computer Vision to scan the video feed of the surrounding to detect any signs of danger. The system can also detect dangerous objects like guns and violent weapons, acts of violence and signs of danger. The system is not limited to home intrusions; it can also keep an eye on the nanny to make sure she’s not acting violently with the kids, and call you on first signs of abuse. A nanny for the nanny!

Using Multimedia Fusion, where both the video and audio feed of the surrounding are analysed at the same time, the security system can detect signs of violence and intrusions even if the action takes place out of the view of the camera.


How does this work?

No Internet Required

While conventional CCTV and cameras stream the video to a central location for viewing and analysis, IRA’s smart systems have the detection algorithm right on the device. The product uses Edge Computing to remove the dependencies of having an internet connection for the detection process. This means that the data from the cameras or phones don’t leave your device. This is a good measure which ensures the user’s privacy is maintained.

Low Power Consumption

Arik from Wise explained that the most compute-intensive stage in the process is the actual AI training process. The initial training for AI component is done at Wise Systech’s servers, where the system is fed with videos, images and audio of danger signs that their systems need to detect. This Deep Learning process provides the system with enough information which is used to create a detection algorithm or program which is then optimised for installation on the smartphone or camera.

This optimised software allows the smartphone or security device to run in a non-intrusive ‘half awake’ state until danger is detected. Only when this situation is detected, does the application wake up and trigger the required actions.  This helps keep the power consumption of the device or phone low while it runs in the background.

Let Artificial Intelligence Help You

This product shows a lot of promise in the helping us in times of trouble. We in India need all the help we can get to keep our surrounds and people safe. With so many incidents of rape and violence reported daily, if products IRA can help in getting help early in case of emergencies.

Keep an eye out, IRA Voice & IRA Visual will be available on the market starting May 2018. IRA Voice will be available for all Android smartphones in the initial roll-out and the IOS version to follow shortly after.

Here are the links to the demo videos of IRA Visual & IRA Voice from Wise Systech.


Share this article
Prev Post

Cut through the SMS junk with SMS Organizer

Next Post

Limiting Tech and Gadgets for Kids

Leave a Reply

Your email address will not be published. Required fields are marked *

Read next