Sleep Hacking, Night 1

Monday, September 26, 2011

There's a new gadget in my life, and it's like a Garmin for sleep. It's called a "Zeo sleep manager", and it's a fairly pricey, extremely dorky, and fantastic piece of technology.

I've always been bad at sleeping. I seem to require complete silence and darkness, perfect temperatures, no smells, and comfortable bedding; basically, I need sensory isolation, which is tough to achieve. As a result, I have a bad habit of lying awake at night tossing and turning, then sleeping late in the morning. As a grad student, I can usually get away with this, but TAing an 8 am lab last semester really convinced me that I need to start adapting to how the rest of the world sleeps. I'd also like to be able to survive ski trips and such without feeling dead after a night of staring at the hotel room ceiling.

I can't remember when I first read about the Zeo, but I was immediately intrigued. I admit to having a certain obsession with quantification, which is why I like the concept of things like MAF tests for tracking improvement in running. The Zeo promised the ability to track not only the quantity of sleep, but the quality thereof, with only one downside: a $400 price tag. However, I continued to read about it and check prices from time to time, and a couple of weeks ago after yet another sleep-deprived night, I pulled the trigger and bought one off of Ebay. The day it arrived, I was excited to go to sleep (which, as my roommate pointed out, is rather counterproductive) and start measuring brainwaves.

The Zeo comes with an innocent-looking alarm clock and a headband that looks like a softer version of a heart rate monitor. Once that band is strapped to your forehead, it starts recording EEG signals through three silver-cloth electrodes. It's been validated against polysomnography and found to be approximately 75% as effective at determining the sleep stage, and 90% as effective at distinguishing between sleep and wake. That's pretty good for a relatively inexpensive consumer device that's wireless, comfortable, and comes in shiny packaging. In any case, let's see some data:

My first night sleeping with Zeo

Pretty nifty eh? This graph shows that I lay awake for 47 minutes prior to sleeping, which is about average for me. However, it also says that my 7 hours of sleep were pretty great, with a lot of deep sleep and a decent amount of REM. It's also interesting to see that I didn't get much "useful" sleep after my 6 am wake-up (I think it was the cat), so perhaps I should just start cutting my nights shorter.

There are a few interesting discrepancies between my perception of the night and the graph. Most significantly, I didn't remember falling asleep and waking up twice between 12:30 and 1:15; I thought I was awake the whole time. I'm guessing the sleep stages during that time (REM, light, then deep) aren't quite accurate, but I'm inclined to believe that I was asleep despite my perception. Similarly, I thought that 6 am wake-up lasted a lot longer. I've been reading a bit about sleep psychology lately, and apparently a big problem with insomniacs is that they think they're awake, when really they're asleep. Could it be that all this time I've actually been sleeping fairly well, but tricking myself into believing that I'm awake? I don't think one night of data is going to tell me the answer, but it's an intriguing idea. Perhaps Zeo will help me by simply giving me more confidence that I got sufficient sleep.

Another way of measuring sleep that's gained popularity in recent years is using the accelerometers built into smartphones. Before getting the Zeo I used Sleep as an Droid, the Android app with infuriating grammar. This uses the principles of actigraphy to correlate motion with sleep stages. However, Sleep as an Droid doesn't try to quantify the amount of time spent in each stage; it simply provides you with a smoothed graph to show your motion throughout the night. This graph, while giving me a bit of an idea of how the night went, doesn't provide data for long-term trends, as it simply assumes you're asleep the whole night. As a result, a terrible night where I lay in bed for six hours and slept (so I thought) for 3 would be reported as a fabulous 9 hours, which was even more depressing than not having any data at all.

In order to compare the Sleep as an Droid data to the Zeo data more effectively, I made a python script to read the raw data, average every five minutes, and plot the result against time. This gives something slightly easier to compare than the previous image:

Sleep as an Droid reformatted plot

Really, I should display these two plots side-by-side, but I haven't quite figured out Zeo's raw data capabilities yet. In any case, I noticed a few things:

  • The accelerometer data between 2:15 and 3:15 (my long period of deep sleep) doesn't look special.
  • My 6 am wake-up appears to be present, but is barely above the noise threshold.
  • My second period of deep sleep (just before 5 am) shows even more motion than most of the night.
  • I can't interpret the Sleep as an Droid graphs.

From these observations, it's clear that there's a discrepancy, and that Zeo provides more useful information. However, it is undeniable that I either move during deep sleep, or Zeo incorrectly detects deep sleep when I'm really in a period of lighter sleep. As delta waves seem like the easiest to pick up on from a signal processing perspective, I'm inclined to believe Zeo when it says I'm in deep sleep.

It's rather unfortunate that I bought Zeo just a week before a cheaper and more convenient mobile system was released, but I don't regret it. I'm looking forward to using Zeo long-term to track my habits, increase my ability to perceive my sleeping, and hopefully improve my ability to sleep like a normal person.

Zeo Personal Sleep Coach: $15 off + Free Shipping


  1. Interesting!
    I've used several of the binaural beats / brainwave entrainment programs to aid getting to sleep and to enhance lucid dreaming. I find it works, but I've also done meditation so my mind is prepared to "fall into" relaxed brainwave states. The software just makes it easier.
    It would be really interesting to combine that with data from the zeo, and get a better idea of what's actually happening.
    BTW, have you ever heard of a computer game called "Wild Divine"? Uses a biofeedback device instead of a mouse. I played with it when I was at Sun. Unfortunately the price hasn't dropped in the last 10 years (and it seems that Deepak Chopra has taken it over - the two may be connected).

  2. I tried a binaural beats program once, but since I don't like sound when I'm trying to sleep it just felt like it kept me awake. However, I've also never meditated, so perhaps it's a mental state that my body just doesn't understand.

    There's a fair bit of chatter on the net about using the Zeo as a tool for lucid dreaming. I could definitely see it having some potential, and since they released an API to access the data, I think there's even some people working on specific lucid dreaming hacks.

    Wild Divine looks interesting, but you're right - crazy expensive! It all looks a little "gimmicky", but then again, so does Zeo.

  3. Very cool!... I wanna try it!.. strap it on before my next meeting and see if I really do sleep!

  4. I've always had similar issues to the ones you describe in this article. I sometimes fall asleep shortly, but then somehow immediately wake myself (your graph suggests diving into REM right after lights out... ), and/or don't realize that I was actually asleep to begin with. And, I've always wondered about the quality of my sleep, seeing as sometimes I wake up more tired than usual. At $99, the mobile version seems like pretty good'd be an interesting experiment. Can the mobile data be downloaded & correlated as well...was it their online tools that you used to get your hands on the data?

    "Analyze your sleep patterns using Zeo’s mobile app, online tools and expert sleep coaching program"

    Again, great article...

  5. p.s. the referral link is broken (Coldfusion stack dump)

  6. Denis: I was just about to reply to your comment, then I noticed the broken link! Apparently I had to re-register or something, and this time it let me select the form of my referral reward - hopefully it works.

    The mobile version seems to be identical with the exception of the alarm clock. In fact, it's even more convenient because I think it syncs the data with the myzeo server through your phone, whereas with the base station you need to pull the SD card out. However, there's a "raw firmware" that can be put on the base station to play with the raw data a bit more, so I suppose there's an advantage to the "bedside" model. I agree though - the mobile one looks to be great value, and I wish it'd been announced before I bought the other one. However, there was no way to know, and this way I get to play with it right away,

    Thanks for the compliment, again!

  7. Sounds the headband comfortable or is it annoying when you're trying to fall asleep?

  8. Janine: The headband is pretty soft. I don't have a problem with it, though I recently realized I had it on too loose (out of fear that it would give me a headache). I wear earplugs pretty frequently though, so I'm used to having foreign objects around my head while I sleep.