Team Members

Industrial accidents due to drowsiness and stress used to be a part of the job. A necessary risk, but not anymore.

Our project's aim is to reduce this risk, using EEG electrodes and IMU sensors by tracking the worker's state of mind and activity we can predict drowsiness. We alert the user and their supervisor about their condition.

Many studies have shown that sleep deprivation can increase the risk of accidents by 70 %! " And tragically, in one Swedish study of nearly 50,000 people, those with sleep problems were nearly twice as likely to die in a work-related accident. "

Source : https://sleepfoundation.org/excessivesleepiness/content/the-relationship-between-sleep-and-industrial-accidents

How our helmet works :

Before we continue, it would be advised to look at the attached image of the setup. The headband consists of EEG electrodes placed comfortably on the user's forehead. EEG electrodes continually monitor brain wave activity of the user. This is processed by an onboard Micro Controller which then evaluates the condition of the person. If the output comes out as sleep deprived, a notification is sent to the manager and surroundings workers via phone to evaluate to take suitable action.

What have we done :

With 0 data sets available online which could help us with accomplishing this task, we, in fact, created our own datasets. That's right, we actually became sleep deprived for weeks together and managed to build the data set, train a Neural Network and get accuracies as high as 98% on the testing data. This was great, but one slight drawback we faced was the fact that our processor, the Raspberry Pi, would be quite bulky to place in the helmet. Thus, to not compromise on the helmet's normal structure, we shifted the processing to a local server. We realize that for scaling up, this would be quite the problem, and are currently in pursuit of hardware capable of doing real-time processing on a small enough form factor to be fitted inside the helmet.

Why is it innovative?

1. It's a retrofit onto existing helmet designs to check for drowsiness in people.
2. With the help of the IMU sensor, the helmet has fall detection capability.

How would it be produced?

At present, the cost of the entire setup (including helmet) is approximately $25. This will surely come down once we design the PCB to replace all of the existing breakout boards ( they occupy much of the board's space).

Potential Impact:

Besides saving lives which are lost due to the "rat-race" life most workers lead nowadays, there are multiple opportunities which present themselves in making the helmet "smart". The following are ideas ( not just top of the hat, based on my team's experience in this field )

1.Health analysis of individual workers(besides just sleep deprivation, including respiratory analysis based on EEG processing, Mental state analysis, if they are under the influence of alcohol, drugs .. just a few to include)
2.Location of the user
3.Interactive long-distance communication

A full-length ppt of technical aspects of this project can be found here, along with the paper which got accepted at the Industrial IoT Workshop, Leuven, Belgium. :
https://drive.google.com/open?id=1M1gpTngarg-i7NDvdRHQZirBEvAHwuEC

Our current short-comings:

1. An expansive data set of multiple age groups would be required to affirm our studies. Thus, we released our paper to have other like-minded researchers to test out our theories and make this idea a reality.
2. Wireless Charging for the helmet: A possibility we are looking into, as it would be convenient, but at present very low efficiency.
3. A suitable processor for onboard processing; we are facing multiple challenges in this domain so any help would be much appreciated.

 

 

Inspiration

Near our college, there are ongoing construction activities which extend past midnight. Looking at the condition of workers and seeing many of them fall asleep near the construction zone itself, we realized that there could be a serious problem if someone were to fall asleep on the job. After exploring into the possibilities and testing out Brain wave analysis, Facial Recognition and Heart Rate analysis, we concluded that Brain Wave analysis would serve the purpose quite well. So, we pursued on this arduous journey to bring the results we discuss today.

Voting

Voting is closed!