I’ve been snooping around here for a while but today is the day I will begin to participate. A little about me: I’m finishing up a master’s degree in electrical engineering and I am the lead student researcher at the dLUX Light Lab, which looks at how various environmental factors (but especially light) affect our health. We’re primarily concerned with the color and intensity of light at different times of day and how these factors affect our circadian rhythm.
For my master’s thesis, I’m designing a light control algorithm that tracks an individual’s sleep quality (via an actigram) and reaction timing (PVT and Seth Robert’s Brain Tracking program) and tries to optimize the light output of Philips Hue lights at different times to improve these measures. I’ll be testing the first version on myself but will also invite others to participate so we can refine the algorithm. More on that later.
Speaking of lights (my obsession), I’m working on an automated circadian light control system that interfaces with multiple companies’ tunable lights and motorized blackout shades. It will automatically mimic the sun over a pre-defined schedule that is adjustable, and as time goes on, it will also get more adaptive features like the algorithm I described above. This is something that should be ready for market in 6-8 months. I’ll talk about this more in a separate post, but would anyone be interested in beta-testing this system?
Finally, the methods we use to track and improve our own health are getting better and better with time, but it seems that taking a big data approach to the quantified self, i.e. combining everyone’s (anonymous) QS data into a huge, open database, would have numerous implications for medicine and society. I’m sure others have thought of this, but I am unaware of any projects that are actually trying to aggregate the data. This is something I would like to start/participate in, but I definitely need some assistance. More on that in another post.
I hope that’s a satisfactory introduction! Talk to you all very shortly.
There are a few services that let you “donate” your personal data for research, and companies like Fitbit generously share your data with “research partners”, but there’s not all that much you can do with a bit of superficial data like step counts.
There’s also the 100K Wellness Project, which aims to collect enough kinds of data to be useful; will be interesting to follow.
I know it’s still a touchy subject for most of the companies, but the main idea would be to get the raw time-series accelerometer/heart rate/etc. data from Fitbit, Basis, etc, then probably run some normalizations to account for differences across manufacturers. That way, you’d be able to look at large data sets and learn about how self-identified populations act under a variety of conditions (like the cool visualizations Jawbone has been putting up lately). Perhaps with enough peer pressure, they’ll be willing to share (or not).
The 100K Wellness Project looks really exciting - I can’t wait to hear from them once they start seeing results!
These companies are more than willing to “share” (for publicity or $$$), since their valuations are based on the assumption that their data will prove to be valuable
I don’t think any of these services collect raw accelerometer data. They do have special research versions of their devices that can capture that data, but battery life ends up being <1d when you do that.
You’re right. I think even the research-grade devices (such as the Philips Actiwatch) don’t export the raw data but an average over some time period, such as 1 minute. That’s how they can achieve a battery life of about 1 month and be very minimally intrusive. The time-averaged values are probably good enough, as long as the periods are not too large.