Vital Master

• 

  ui0

• 

  ui1

Inspiration

Healthcare is a very important topic for everyone. Many diseases can be diagnosed before serious problem happens. For example, the body movement data can provide very rich information about the health status. A freezing gait while walking may indicate early Parkinson's disease. Vital signals, such irregular breathing or heartbeat may also indicate health problems. There's an urgent need for ubiquitous, continuous vital signal monitoring so that serious diseases can be predicted and prevented. Besides, we believe such long term data, if collected properly, will make a big difference in healthcare in this big data and AI era.

What it does

We built a system that monitors/collects human 3D body movements, respiration, and heartbeats, without attaching any device on the body. Such data can be used for real-time emergence notification (e.g., when people falls on ground or has a heart attack).

How we built it

Our system consists of two hardware pieces, one XBox Kinect sensor and one radio chip (like Wi-Fi router). We track human body joints from Kinect depth camera. The radio chip sends out radio waves which get reflected from human body. Body movements such as respiration and heartbeat, will have impact on the signal patterns, thus making it possible to be estimated from received signals. We segment out reflection signal for each user, and estimate the respiration rate and heartbeat for each user.

Challenges we ran into

1. We used two kinds of hardware: Kinect and radio chip, which require their SDKs or python libraries. We met some compatibility issues to set up both in one environment, which takes a lot time.
2. In the software design, we need to use two data streams from Kinect and radio. They need to exchange and share data, sharing data among multiple threads takes some time to solve.
3. This is the most challenging part: radio signal processing for vital signal extraction. We spent a lot of time trying different algorithms and did a lot of testing experiments. Due to the very limited time, our robustness, latency still need improvement, and some features need future work.

Accomplishments that we're proud of

We are dealing with a BIG problem for everyone! We wish our technique can benefit the patients in the hospital, the old living alone in the home, the babies sleeping out of sight, and basically everyone who wishes a healthy life! We are also proud of the difference our technique may bring to AI health. Real data is precious for AI!

What we learned

We learnt how to track multiple human movements using Kinect sensor, and how to measure human body vital signals using radio chip. Now we are more familiar with python development and signal processing techniques.

What's next for Vital Master

We will add face recognition feature into it so that we can label the data with user's ID. We will also keep improving the vital signal monitoring robustness and finish heartbeat detection which is currently half done.

Built With

• c
• kinect
• python
• radio