Asthma is one of the major diseases that my grandmother has faced. Because of which she can’t put on a face mask for the longer duration of time. She always wishes to have something that can perform both jobs parallelly, protect her from coronavirus and also not suffer from difficulty in breathing. That is why this Idea of making a face alert system pitched to me.
What it does
An app that works like an alert mechanism to prevent us from touching our face and hence it can help to prevent the spread of covid-19.
If the user moves his/her hand towards their face then the app generates an alert and the phone starts vibrating
How I built it
It works as follows:
When the user moves his hand, the device collects accelerometer data and sends it to the API as a JSON file.
The API is backed up with a machine learning model.
The machine learning model is well trained to differentiate when a hand is near the face and when it is not.
Our backend algorithm calculates the Roll, Yaw and Pitch using accelerometer data which can be used to define the orientation of the device.
So whenever the user brings his or her hand near the face, the backend model raises an alert and the device starts vibrating.
We’ve further integrated Watson ChatBot with our app to help people with their covid related queries
Challenges I ran into
Q. What all devices and OS your app has supported? We’ve currently released apps for smartphones (for testing purpose) and smartwatches (for Wear OS watches), development of the app for other devices are under processes such as Fitbit and Fossil.
Q. What if we don’t have an internet connection? The first version of our app requires internet to operate but we’re working on to deploy the system locally and next version of the app won’t require Internet to work on.
Q. While I’m on call will my band vibrate continuously? No, we’ve taken care of that problem therefore when you’re on a call the app gets an alert and it doesn’t vibrate, we’ve achieved this using PROXIMITY SENSOR.
Q. What is the accuracy of the model you’ve used? We’ve used different models to test and the model with the best accuracy is used for deployment. Currently, it’s accuracy is around 91%.
Accomplishments that I'm proud of
We are in top 16 teams of Crack the COVID-19 Crisis (NASSCOM FutureSkills and IBM)
What We learned
We learned how to work in a team and how to handle critical situations
What's next for Face Alert System
Testing of Fitbit app is under process and the stable version will be released soon also were focusing to attain compatibility of our app with other fitness bands as well.
We can add an additional feature in future versions of the app that will alert the user if they’re near the infected user. Data of the users will be stored in the cloud Database Hence this will attract more and more users to use our app
We’re planning to deploy our system completely locally. Hence, users won’t require in to access the services.