What the devil is actigraphy?

Actigraphy is the use of accelerometers (motion trackers) to monitor physical activity, whether waking life or sleep. Pilots, medical professionals, and militaries have used it extensively for decades, and in more recent years increasing numbers of consumer products have been released. The most simple example is a pedometer, and more complex devices include the WakeMate, FitBit, and Jawbone UP. Even some apps will turn your Smartphone into an actigraphy device.

Use of actigraphy for sleep seeded in the 1960s when predictable limb activity was discovered during sleep studies. It was found that wearing an actigraphic device on a person’s non-dominant wrist was actually just about as accurate as more disruptive forms of tracking sleep such as attaching electrodes to your skull. But how is that possible? What is your wrist up to that allows a relatively simple gadget to match the accuracy of machines that measure your brainwaves?

First let’s define the human sleep cycle…

There are 3 stage of sleep that have been tweaked and expanded since being first defined in the 1930s, plus wakefulness (sometimes called stage 0). Briefly:

Stage 1: Drowsy stage; you’re twitching and jerking around, and sometimes this stage feels like hallucination or like you’re half-awake. If you wake during this stage you’ll swear you didn’t sleep at all, but you actually did.

Stage 2: Light sleep stage; your brain is continually slamming on the brakes to keep your mind tranquil, but you can easily be woken in this stage. About 50% of your sleep cycle is here.

Stage 3: Deep sleep stage; this is when you dream, regenerate tissue, it’s difficult to wake you, and REM is interspersed during stage 3. If you wake up groggy, you were probably in the middle of Stage 3 sleep.

Stage 3.1: REM (“rapid eye movement”) is a subset of Stage 3 and the bread and butter of sleep. Quality REM keeps you healthy, creative, and well-rested. Most dreaming occurs during REM and usually you have 3-5 REM sessions per night of increasing duration. About 20% of adult sleep is spent in REM.

From here I’m going to use the terms NREM (non-REM) and REM to discuss sleep as this seems to be how most actigraphy devices report your sleep — though they usually call them “light” and “deep” sleep, respectively. Specifically NREM is Stages 1-3 without the REM parts of Stage 3, and REM is obviously only the REM portions of Stage 3 sleep.

So for actigraphy and sleep…

Jawbone UP feels your pain.

The TL; DR of this post is that large limb movements occur during NREM states and rapid micro-movements take place during REM. Measuring and analyzing these movements with surprisingly simple algorithms is how an actigraph can account for how well you are sleeping. In the vast majority of people wrist micro movements during REM measure around 1 movement per 3 minutes (0.3/min), whereas during NREM a similar movement is 1 movement per ~90 minutes (0.01/min). [Page 151, Activity Measurement in Psychology and Medicine by Warren W. Tryon]

Additionally helpful for your actigraphy device is that during REM sleep your muscle tone is reduced to the point where large movement is not possible (pseudo-paralysis), so if it measures large movements then it knows you are either awake or in NREM. Or you have a sleep disorder where you act out your dreams in your sleep, similar to Disney cartoons where we see dogs’ legs moving to match their dream of chasing a cat.

Lastly sleep activity decreases with age, as well as other factors, so when you tell your WakeMate, UP or Fitbit how old you are potentially this is helping it increase accuracy of sleep tracking. [ibid, p. 153]

Further down the rabbit hole…

We all sleep differently so how does sleep actigraphy account for this? A device can evaluate what your movement behavior has been for each epoch (interval of time, say 60 seconds) over a period, and then use standard deviations to estimate whether you are transitioning stages of sleep or whether your partner bumped you at 3am. If your movements continue to land outside the standard deviation of your normal REM movements it will presume you are awake. In a sense the actigraphy device can customize itself to your personal behavior rather than using any absolute scale. Pretty cool.

WakeMate trend view

I don’t believe any devices have this feature yet but it’s imaginable that if you have large movements for 20 seconds after 45 minutes of normal REM movements that would indicate you are awake, it could perceive that as “noise” and ignore it if you then continue REM behavior afterwards, or it could evaluate that as potential sleep apnea or other sleep disturbances and disorders. I’m completely guessing, but perhaps WakeMate and Jawbone UP use a similar model to determine part of your sleep “score”. Maybe if it records crazy movements throughout REM sleep it may decide that you had a poor quality REM cycle.

In the future actigraphy users hopefully will benefit from software tracking of their sleep cycles over time against pre-slumber activities. In the hours leading up to sleep were you watching TV, reading, exercising, or drinking heavily at another Kiss reunion concert? Do you actually sleep better after taking a sleeping pill (i.e. better REM)? I’ve found intriguing trends with my Jawbone UP by taking notes manually, and look forward to their updates for even better analysis.

To be more specific: out of the box Apple TV only decodes H264 video and has some really old hardware, however by replacing the Wifi card with a brand-new 1-watt hardware decoder called the “Broadcom Crystal HD” you can start playing very high quality video for other codecs such as DivX, Xvid, WMV, and so on. As for a video library, personally I’ve opted to stream video from a NAS in my house, but others have upgraded the internal hard drive on the Apple TV to hold more videos.

For this project you’ll need:

- 1st Gen Apple TV (40GB or 160GB)
- Torx T8 and T10 screwdrivers (to install Crystal HD)
- Broadcom Crystal HD BCM970015 (~$50)
- Non-wifi connectivity for your Apple TV (we’re replacing the wifi card)
- About 45-90 minutes depending on your comfort level

Alright, let’s get started…

1.) Install the Broadcom Crystal HD BCM970015 (or the inferior but okay BCM970012 or Habey HB-VD904). Logic Supply has some as of Oct 2010.

2.) Follow Gizmodo’s well-written guide to hack your Apple TV for Boxee, except on step 6 just install XBMC as well as Boxee (it’s in the same place).

3.) Install custom Crystal HD drivers [source] by running the following commands in the Terminal in Mac OS X (or whatever you want in Windows) [credits]:

ssh frontrow@appletv.local

The default password is “frontrow”. Now enter this command into the terminal once the password is accepted to add a menu item for downloading the drivers:

defaults write com.teamxbmc.xbmclauncher XBMCAdditionalDownloadPlistURLs -array-add http://crystalhd.sartori.at/files/crystalhd.plist

Also add a menu item for the nightly build of XBMC; this makes the very most recent version available rather than just major releases:

defaults write com.teamxbmc.xbmclauncher XBMCAdditionalDownloadPlistURLs -array-add http://mirrors.xbmc.org/nightlies/osx/update_urls.plist

Finally, go to the Launcher menu and then Downloads in the main Apple TV menu and select “Crystal HD r171 (20101009)”. Do the same for the nightly build (e.g. “XBMC svn r33778″). These two downloads will install the driver and update XBMC respectively. Reboot the Apple TV afterwards.

4.) Launch XBMC and confirm the setup worked by navigating to System -> Video -> Playback and the second item down should say “allow hardware acceleration (crystalhd)”.

5.) Profit.

That’s all! Enjoy the 720p and 1080p HDTV video. Btw, (re)focus is a really well-designed skin for XBMC that works more fluidly on the Apple TV.

PS: You can remove the extra download menu items we added in step 4 by using the following Terminal command after SSHing into the Apple TV:

defaults delete com.teamxbmc.xbmclauncher XBMCAdditionalDownloadPlistURLs

These impossible sights were captured by “Doc” Edgerton in 1952 at a distance of 7 miles (11.2km) using his recently invented rapatronic camera. He was an electrical engineer from MIT turned photographer and utilized an interesting technique to avoid slow mechanical parts in a shutter. Edgerton co-founded EG&G based on his work and it currently employs 11,000 for NASA and US defense contracts.

The first photo is approximately 1 millisecond after detonation while the blast diameter was 20m. The exposures are (ostensibly) 1,000,000,000/sec., or 10ns. In the later photos you can see rope effect; effectively vaporization of material traveling down the cables and trellising. Unlike modern high-speed photography only one shot could be taken per camera and Doc set up 10 cameras for this blast. I’m missing six photos…

Let’s hook it up!

24x16 LED Matrix Wiring (made using Fritzing)

If you reverse the GND and 5V wiring you will probably blow up the display. The pin numbers are on the back of the display, so be sure to reference those. The connections:

- Arduino Digital Output 12 to pin 1 of the display connector (CS).
- Digital Output 11 goes to pin 5 of the display (WRCLK)
- Digital Output 10 goes to pin 7 of the display (DATA)
- Digital GND goes to pin 11 of the display (GND)
- Power 5v goes to pin 12 of the display (+5V)

Now to the software. I found a post by westfw (#94) on Arduino.cc with some sample code that goes through a variety of demos on the display using various contributors’ code. I’m a novice programmer at best, but I took his demo and cut out just the core parts for the display and wrote my own implementation to have the whole matrix select and toggle random LEDs on and off. Download the whole thing here. It’s not pretty, but here’s my hasty code:

void setup ()
{
ht1632_setup();
randomSeed(analogRead(0)); //get some analog static from pin 0 for the random number generator
}

void loop ()
{
int x, y, p;

x=random(24); //pick a random X coordinate
y=random(16); //pick a random Y coordinate
p=random(2); //pick on/off (1 or 0, respectively)

plot(x,y,p); //tell the display to execute the randomness

}

So if you’re vaguely familiar with Arudino (like me at this point) you should be able to plug wires in, run the code, and watch some LEDs blink randomly. Behold the awesome power of microcontrollers!

Related links:

- Sure Electronics DE-DP016 Data Sheet
- Arduino.cc Forum Post on the DE-DP016 (and related displays)

What am I getting at? This: a plethora of potentially fly-by-night or unproven backup services have gone live over the last year or two and require a bit of investigation before subscribing. Most of these services will be able to stay afloat for quite a while due to the low overhead, but my warning is to beware of the little guys until they have a track record. There is little investment on their part, often just a few lines of code and the payment system. Usually there is no mention of data security and a lack of welcome features such as having data restore to DVDs and mailed to you.

I’m not saying avoid S3 backup services, I’m recommending confirming there is a history of quality support and updates. JungleDisk is one such candidate. But you might just want to stick with inexpensive non-S3 services such as Mozy, Carbonite, and so forth.

Here’s a list I compiled of some backup titles and services leveraging Amazon S3 storage:

• Backup Manager
• S3 Backup
• S3 Solutions
• Sync2S3
• Brackup
• s3DAV
• Duplicity
• s3sync
• SetS3t Synchronize
• DropBox
• JungleDisk

Happy uploading!

I suspect they are planning collaborative features, but currently its single-user creations. A timetoast “wiki” of sorts would be fantastic.

There wasn’t a timetoast timeline for the iPhone, so I filled in the gap.

(Above I’ve charted the initial cost of the iPhone + setup fees + monthly service over 2 years, including taxes and fees. I’ve added a $5/mo text plan and calculated the totals using California sales tax (8.25%) – so the bottom line in different states may be slightly more or less.)

More after the jump, plus another chart of TCO without a text plan…

The result seen here is that even given the lower initial cost, the iPhone 3G costs more to own than the original iPhone at 14 months and after 2 years is a little under $200 more expensive. This is due to the monthly 3G service plan being about $15 more expensive for unlimited 3G data + 200 texts.

If you are one of the few smartphone users that really, truly never use SMS text messaging, it takes until about month 20 for the cost of a iPhone 3G to surpass the original iPhone. Here the price is much more closely matched, though as mentioned earlier the original iPhone service includes a 200 SMS plan and the iPhone 3G plan has none.

Personally, GPS and 3G make it compelling enough for me to upgrade. Plus I suspect the original iPhones will still be worth a pretty penny (easily $200+) on eBay even after the iPhone 3G is available.

• With a two-year contract, the price of an 8GB iPhone 3G will be $199; the 16GB model will be priced at $299.

• Unlimited iPhone 3G data plans for consumers will be available for $30 a month, in addition to voice plans starting at $39.99 a month.

• Unlimited 3G data plans for business users will be available for $45 a month, in addition to a voice plan.

Additionally, this may pose an issue to upgrading iPhone users, since potentially they won’t be able to take advantage of 3G without entering a new 2-year contract. This is not particularly surprising, particularly given all the recent AT&T 3G infrastructure improvement, but definitely a bit of a kick to the pants. It appears this increase may be across all data phones as well, including Windows Mobile and RIM Blackberry smartphones.

2:42pm Update: Gizmodo confirmed this with AT&T’s Glenn Lurie, plus you must activate the phone in store (Apple or AT&T). Ouch. At least that should keep the resale value of 1st gen iPhones relatively high.

AT&T: AT&T to Offer Next-Generation iPhone on Its High-Performance 3G Network

Similarly, the Atom meets the basic computing requirements of average users but not much more. This means you can check email, browse the web, use Office but can’t utilize advanced HD video, 3D games, and so forth.

An exceptional use for the Atom platform is a home server due to fantastic performance per watt. It uses around 2 watts of power under load, whereas a common processor such as the Core 2 Duo uses 60W+. As energy costs go up, the total cost of ownership of a server is skyrocketing. So while the intial cost of a server or desktop suitable for running as a server may be low ($500 on a budget), at 24/7 the energy cost will be $100-$400/yr. In contrast, a 24/7 Atom-based server would cost approximately $10-20/yr. Substantial savings.

There are some great uses for Atom home servers:

1. BitTorrent. Dedicated 24/7 upload and download without worrying about utility bills.
2. Home automation. You can leave a home-automation management server running 24/7/365 for probably under $7/yr in energy.
3. Centralized storage. You can make your own NAS – most efficient in conjunction with the above services.
4. Music server. Always-on access to locally-stored and Internet music.

Using the Atom platform for a home media server is possible, but only with streaming lower quality video. HD quality isn’t going to work.

TranquilPC offers an Atom-based D945GCLF motherboard for about $100US. Throw RAM, flash-based storage, power supply, and a case on top of that and you’re looking at about $200-$300 total. TranquilPC and other vendors are starting to offer pre-built systems as well, but generally run $500 and up.

Anyone have other ideas for the Atom platform? Inexpensive solar-powered server or desktop, maybe?

Mac and Windows

Right off the bat, we can see that Mac OS holds about 30% of the desktop and laptop OS market and Windows approximately the other 70%. I have left Linux out for now (see below) because it is nearly impossible to determine whether it is being used for a desktop computer or server.

Windows

If we explode the 70% Windows market share, things start to get a little more interesting. We can see that Windows XP commands 79% Windows market share. Vista has only been able to penetrate 10% of the market, barely statistically greater than the 8% being possessed by 8-year-old Windows 2000.

It’s also notable that Vista has about 10% market share despite something like 1/3rd of computers on Stanford campus having been purchased since the release of Vista. That means the vast majority of new computers are being ordered with Windows XP. Clearly Vista’s market penetration leaves much to be desired. Additionally, 10% is actually fairly representative of the entire US penetration of Vista, through the causes may be different (Stanford users have a higher average turn-over rate on computers than the rest of the US).

I could explode the particular flavors of Windows once again into each individual edition (Vista Business, Enterprise, Ultimate, etc.), but I’ll just mention that Windows XP Profession holds about 95% of XP editions and Business is the majority of Vista. Usage of “Home” editions of Windows is discouraged at Stanford University by being largely unsupported by the internal IT infrastructure.

Linux

As I mentioned earlier, I chose not to include Linux in the first “desktop/laptop” OS market share because it is nearly impossible to differentiate usage of Linux for servers versus desktops without enormous statistical effort. However, if you include all version of Linux, about 16% of computers (including servers) at Stanford are running Linux – a large portion are email, storage, web, and various other servers.

Linux flavors with very low representation (e.g. Gentoo’s nine installations at Stanford) are excluded from the list. And unfortunately, nearly 1/3rd of Linux instances are reported generically as “Linux”. Lazy Linux users…

Final notes

Stanford is a great look at a free-market edu computing environment with strong Mac and Linux usership. But perhaps most importantly, its low Vista penetration shows that average educational users, whether student, faculty or staff, are still very resistant to Vista and go to extra effort to purchase XP even with new computers. Vendors, such as Sony, choosing to move away from XP earlier than most may be less attractive in this sort of environment.

In the future I may do a report on computer vendor market share (Apple vs. Dell vs. Lenovo vs. HP vs etc.).