A Key Team
Questions are always our impetus to inspiration. Who, What and How drives us. Who else can be studied and attributed for your personal well being other than you? What can be done well - in terms of cost effectiveness and time efficiency to us, with something you ought to use everyday - Laptops, Computers ? How can we make it extremely simple and non-intrusive, in order to reach the masses?
What it does
Q-KEY - Quick Key to health
The plug and play product works by detecting your vibration pattern while typing in an unobtrusive way on a device that you already own and classifies the data into your emotional states. Based on the level of your stress states classified by our complex algorithm, the application will pop-up and try to suggest some preventive care methods. If the data consistently shows that if the user is heavily stressed, then the need for the preventive hospital care arises. Through a simple pop-up, the user will record his current irregularities and if the doctor's assistant is available for a call, he/she can make a call and schedule a appointment with the doctor for further diagnosis, if it is worse.
How we built it
We used a Piezo Electric vibration sensor coupled with the Arduino (Which can be modeled as an USB drive in future) attached to the Laptop/PC. This we call as QKey. Qkey with the help of generated electric charge when they’re stressed due to the vibrations as a result of the typing behavior. This data is then pre-processed to remove the noise. Inspired by the music comparison algorithms, we finger-printed the subsequences of the vibration data and used its Fourier-transform as its feature space. This feature space is then used to classify the data into different stress levels. With this we used Random Forest classifier to be our algorithm for the supervised classification of the limited data we have with many class variables instead of the Support Vector Machine which is limited for this data-set.
Challenges we ran into
Documentation of the Piezo electric sensor - None Hardware constraints: NO Pressure Transducers Received damaged vibration sensor
Feature Selection Finding a Viable Use-case - DONE
Accomplishments that we're proud of
Taking care of many challenges into one solution Supporting mental health through preventive care solution
What we learned
Great Team work Diversified Cultural experiences Of course, a lot of coding and Hardware hacks How to manage high intense eating and working simultaneously :P Machine Learning Best Practices
What's next for Qkey
Integration with the other auxiliary inputs like fit-bit etc - to improve the the model accuracy