Met these folks at the recent Product Hunt meetup and one of the founders showed me a demo of SleepWake which I thought looked great. It sounds like the algorithm behind it is really advanced - it can detect how awake you are based on your heart rate.
Was surprised to find out they weren't on PH! They asked me to post, so here we are.
@bballinger I'm the maker. I'm trying to make my answer using laymen's term, and provide link for jargons :)
The algorithm works like this: the heart rate patterns (a series of beat-per-minute values, like 70,80,88... ) are transferred from the watch to the phone, then sent to the server. At the server, the signal is transformed to frequency domain (https://en.wikipedia.org/wiki/Fr...) and features are extracted, then the features are put through the neural network (https://en.wikipedia.org/wiki/Ar...) which has already been trained before on a large number of people. Then the output (currently has 3 categories: alert, drowsy, asleep) will be returned to the app.
The training data for alert and asleep is taken quite easily, according to normal states when people are 2-10 hours into their day (for alert data) and when people go to bed and have deep sleep (for asleep). The data for drowsy is taken like this. People are required to drive on a simulator when they have already stayed up 20 hours non-stop. They drive until they crash repeatedly on the simulator, regardless of just being woken up. The data 5 minutes before such crashes are labeled as 'drowsy'.
Hope I answer you well, feel free to shoot more questions.
As of publication, I wish I could tell you which exact papers our team follow, but they are among this list http://ieeexplore.ieee.org/searc...
Cheers.
I like the general idea although I wouldn't use it to see if I should drive or not. I would use it as a way to see if I should nap or drink coffee and I'd also use it as a way to keep awake/alert at a meeting etc. Plans for a version that works for iOS?
@daveparkhere Yes your use case suits the app well. There's a background process running and taking your heart rate at short intervals. Should it recognize you are sleepy, it'll then alert you by triggering the phone alarm and vibrating the watch.
For AppleWatch, as long as Apple allows developer to access to sensor SDK, we'll need a month to make the app. They said they'll do it soon at WWDC, but still not yet. Let's hope the wait is not too long.
Chargedesk
CareerGPT.ai
Cardiogram
CareerGPT.ai
CareerGPT.ai