Inspiration
Family members are always worried about you. They try to call you and you may not pick up, this actually increases their concerns and worries. To solve this problem, we introduced a tool which will inform your loved ones if you are unconscious.
How it works
We collect your daily activity and store it in learning engine. Once, we have some good knowledge about your activity stream, we kick into monitor and predict mode. If we find some consecutive mismatch in the prediction data and current monitored data, we trigger an alarm. If you are conscious, you put it off if not, we'll timeout and will alert your loved ones.
Challenges I ran into
Every user has a different behaviour. So, to combine it and predict it was challenging. We first thought of using Google's Prediction API, but it didn't suited our use case and the API usage limits were also low.
Accomplishments that I'm proud of
We built our own activity monitoring system that learns and predicts the events. This can be used with any sensor to learn its activity.
What I learned
How a learning engine works and how can we make a simple prediction algo based on averages or averages of squares.
What's next for AVENGERS
The prediction algo can be improved further. Push notifications can be sent out with the map location.
Built With
- android
- dropwizard
- express.js
- mobile-sensors-accelometer
- mysql
- node.js
- twitter-digit
- volley
Log in or sign up for Devpost to join the conversation.