Gait Analysis from Satellite

Ignoring the sensationalist headline, this is interesting:

By analysing the movements of human shadows in aerial and satellite footage, JPL engineer Adrian Stoica says it should be possible to identify people from the way they walk -- a technique called gait analysis, whose power lies in the fact that a person's walking style is very hard to disguise.

Video taken from above shows only people's heads and shoulders, which makes measuring the characteristic length and rhythm of a person's stride impossible. That's not true of shadows, though, Stoica told a security conference in Edinburgh, UK, last month. Shadows, he says, provide enough gait data to deduce a positive ID. To prove it, he has written software that recognises human movement in aerial and satellite video footage. It isolates moving shadows and uses data on the time of day and the camera angle to correct shadows if they are elongated or foreshortened. Regular gait analysis is then applied to identify people. In tests on footage shot from the sixth floor of a building, Stoica says his software was indeed able to extract useful gait data.

The article goes on to say that using satellite images would be harder, but that the basic idea is the same.

Of course, this is less useful for finding individuals and more useful for tracking a population as it moves about its day. But some individuals will have more distinctive gaits than others, and will be easier to track. Soon we may all need to walk with rocks in our shoes.

Posted on September 9, 2008 at 12:22 PM • 47 Comments

Comments

Nomen PublicusSeptember 9, 2008 12:35 PM

"Regular gait analysis is then applied to identify people."

Really? This works "in the wild"? That's news to me.

Isn't this like looking for suicide bombers by targeting the people at the airport who look nervous?

RoySeptember 9, 2008 1:03 PM

This is conceivable for aerial photography, but not for orbital photography.

From orbit, looking down through a column of air with a density gradient moving from a vacuum to sea level pressure is equivalent to looking through five miles of air at sea level. The total mass of the air in either column will be the same.

Clive RobinsonSeptember 9, 2008 1:40 PM

@ Roy,

Not only the density gradient but issues with movment and other problems such as humidity effects as well, giving the reverse of "Twinkle Twinkle little star".

Also just how big a radius on the mirror is required to resolve to less than 10cm (1/2 width of shadow from legs).

My napkin calculation for 500KM up suggests there are going to be some significant engineering problems.

sooth sayerSeptember 9, 2008 1:46 PM

Models use their catwalk gait
Politicians are hard to tell apart as they shuffle
Ambulance chasing lawyers run after the client the same way .. (some hide in toilets .. but that's another story)
Rappers swing their hips and arms alike

that covers half the populations -- who are we going to tell apart .. oie veh

Fred PSeptember 9, 2008 1:48 PM

Cool. I hope everyone starts using this, ideally as their only mechanism of identification. It's really easy to learn to change your gait, and for that matter I have different gaits I use for different purposes.

Scott KSeptember 9, 2008 1:57 PM

Another problem is the database--sure, you can verify that Subect A was Locations X and Y, but without first studying and cataloguing a known subject's patterns, you have no way of identifying Subject A.

Patrick OgenstadSeptember 9, 2008 2:00 PM

Let's just hope its sunny when the terrorists roam about. It would be a shame if someone spent millions on something like this only too find out that it didn't work if it was cloudy. :)

Carlo GrazianiSeptember 9, 2008 2:02 PM

This doesn't sound like a useful application of satellite imagery. Low-earth orbit is a 90-minute orbital period, giving minimal loitering time over any target. Plus, if you want surveillance at high latitudes (more than 10-20 degrees, say) , you need a higher-inclination orbit, and you get fewer usable passes over the target (as the Earth turns under the satellite), like 2 to 4 passes per day, probably about 5 minutes each.

So satellite surveillance works well for static targets like buildings, but chances of catching a particular transient target are really minimal.

denis biderSeptember 9, 2008 2:03 PM

Satellites don't seem very useful for this, as they fly by so fast they can only look at a target for a couple minutes every day.

This could however probably be useful to identify people in recordings taken by automated spy planes.

2change2oldschoolSeptember 9, 2008 2:21 PM

Countermeasures include:
Custom shoes, custom pants with length of gait adjusters
Electrically operated muscle control at parts of gait. Surgery to change legs.
Etc. Why spend lots of money on false images, that the sophisticated, that need to evade, will evade, and cloak as others?
This might be great for google, but old school, trained people in the game are the network...
Another $$$ toy that fit the cold war is over, avoid people risk, screwup mentality.

zomgSeptember 9, 2008 2:52 PM

Good lord. Can't people just leave us alone? No wonder people flip out and start shooting! You can't even walk around without being identified by your own shadow.

Doghouse SecuritySeptember 9, 2008 3:07 PM

By now, you've heard that THEY spy on everyone by looking at the gait from your shadow. You know that parasols and umbrellas aren't always effective at blocking THEIR cameras, particularly when THEY have so many trained at YOU, so take:

The Doghouse Security Gait-Altering Course!

Yes, with only 4 weeks of study with one of our Gait experts, you can learn to change your Gait! Fool THEM when they spy on you to thinking that you're MULTIPLE different people, or someone ELSE entirely!

And all this for the cheap price of 500$/week! From a name you can trust!

"They belong in the doghouse!" - Bruce S., security expert, overheard after reading our website.

Click HERE to sign up today!

(minor legal note: offer not available in any jurisdiction with Title 15 U.S.C. §§ 41-51 section 5 or similar legislation is enforceable).

GCU Prosthetic ConscienceSeptember 9, 2008 3:25 PM

Of course, if you use rocks-in-your-shoes to counter gait recognition, you must ensure an adequate supply of randomness in selecting the rocks.

FremenSeptember 9, 2008 3:50 PM

Just take a few lessons from the Fremen in Frank Herbert's Dune series. You'll soon learn to walk in such a way that mimicks desert animals... Feed that into your gait tracker!

"OMG! We've got a swarm of scorpians in Central Park!"

PhillipSeptember 9, 2008 4:08 PM

High-Day.

High day occurs between noon and 2 PM in most places. It's where your shadow is shortest. Within the tropics, there is no shadow at the day's peak, it's directly underneath said person.

What then?

Interesting idea, I bet it'll be MUCH more prone to error than facial, retinal, or other biometric systems.

KentSeptember 9, 2008 4:08 PM

This article about gait analysis says that one can recognize an individual from aerial footage by the way his shadow moves. Bob Woodward was on Fresh Air this afternoon talking about his fourth book on the Bush White House and he said he couldn't discuss "new technologies used to kill specific individuals in Iraq" but said that military people told him we were killing all the insurgent leaders using new secret technology. Gait analysis on unmanned aerial video surveillance? hmmm. Seems like something DARPA could do.

NickSeptember 9, 2008 4:45 PM

Interestingly enough, I believe I recall seeing somewhere one of the most annoying issues with Gait Analysis was "Flip Flops" due to them forcing your feet to actually grip the surface you're walking on.

Quick. Everyone get your flip flops!

Knowler LongcloakSeptember 9, 2008 5:26 PM

@Bruce Schneier
>Of course, this is less useful for finding individuals and more useful for
>tracking a population as it moves about its day. But some individuals
>will have more distinctive gaits than others, and will be easier
>to track.

Why does this matter?

If this is more of a wholesale surveillance type of thing, isn't it the kind that gives too much data to be worth even looking at? Even if you had ideal conditions where you can isolate a single shadow without distortion (which are unlikely in New York at 5:15 pm) wouldn't you also have a false positive rate so high that it is unusable. Also isn't deploying an aerial drone and keeping it up in the air more expensive than a network of CCTV camera's tapes that no one looks at because of the same reasons?

If it is more of a track a single person sort of a thing, then isn't the cost of tracking the person so high as opposed to other means that it is not going to happen except for very specific cases. (Think intelligence operative heading for a dead drop.)

The only danger I see is that public money may be wasted on investing in such technology.

elizillaSeptember 9, 2008 5:35 PM

Since most Americans never leave their homes except in their cars, this will only work on foreign nationals who live in places where people actually walk outdoors. ;)

KashmarekSeptember 9, 2008 7:31 PM

What shadows? The ones on the ground? What if your shadow is in the shadows? Or, if it is cloudy? Walk next to someone, where is your shadow then? They might create an electronic trail of the gait, but who's gait is it? Outrageous security theater.

Gene Wicker JrSeptember 9, 2008 8:31 PM

I would think it would be easier to trace a man than a women simply because women wear all sorts of different shoes, particularly shoes with different heel heights that would affect the way they step.

Davi OttenheimerSeptember 10, 2008 12:24 AM

I can see how normal gait analysis helps people with ailments, and thus justifies the expense of development.

Mitigating health risks makes sense to me and I expect more gait analysis in medical research.

However, this satellite example sounds like a solution still looking for a problem.

Anyone who does not want to be recognized could easily avoid a satellite system, let alone radar.

Shadows from a long dress or a long coat, for example, would render the system useless, no? What about from shoes with wheels in the heels? Terrain also matters. The systems probably assume a hard surface like pavement. Also, it has been mentioned many times here and elsewhere that a simple change in shoes can alter gait.

I did some digging into medical research recently and found that flip-flops cause shorter steps with less vertical force in the heel. Studies show the toes do not rise as much during the leg’s swing phase, resulting in a larger ankle angle and shorter stride length. It layman's terms, we grip the flip-flops with our toes and throw off our normal gait.

http://education.auburn.edu/news/2008/june/flipflop.html

I wonder if these scientists realized they were documenting a major flaw in gait analysis systems.

BWSeptember 10, 2008 12:33 AM

Who needs rocks when they can wear heals or a long dress? Or you could adopt a funny walk for being covert... didn't they do that in Dune to avoid drawing the interest of the worms?

dxSeptember 10, 2008 1:02 AM

I think this is a bunch of crap. However, the thing about the atmosphere doesn't make it impossible. Observatories use lasers to cancel out atmospheric effects.

neillSeptember 10, 2008 4:41 AM

one could make shoes with electrodes controlled by chips, that cause random muscle movement to alter your walk (e.g. bionic man)

wonder if any of this could be admissible in court as evidence one day?

possible that some kind of NASA-esque remote sensing would generate even more acurate data ...

SteveJSeptember 10, 2008 6:16 AM

As a Brit, I'm not very scared of this one. If the British government wants to track me, it would be cheaper and more reliable (think night-time, and cloudy days) to do so using the CCTV on every street corner than using satellites.

If a foreign power wants to track me about my daily (but not nightly, and not when it's cloudy, which feels like at least 80% of the time, and only when a satellite happens to have a view) business, then I guess it can have a go.

I wonder whether there comes a point when some nation decides that another nation has too many spy satellites pointed at it, and starts shooting them down.

bobSeptember 10, 2008 7:17 AM

You're right - satellites are too unwieldy for this.

Lets establish a special security sales tax of 5%, call it the "super-turbo-wonderful-gleefully-willing-patriot donation", added to everything sold in the US, in order to purchase a fleet of 1,000,000 computer operated radio-controlled helicopters (from China) linked to the NSA which can hover at 1,000 feet and video everyone in the country in real time. THEN we'll REALLY be secure!

Of course, because its not 100% reliable we'll also have to wear mortarboards with our social security number printed on top (in lieu of "Thanks Mom&Dad!") just to make sure they can identify us. So this will only work for tracking law abiding citizens and not for people who commit fraud or other crimes. Or terrorists.

EponymousSeptember 10, 2008 9:21 AM

You can change your gait with postural training. For instance, if you have lordosis, you can fix that with weight training. Google for "Neanderthal No More" for one example.

Fred PSeptember 10, 2008 10:06 AM

@Eponymous-

It's not just that - I mean, at least I have very different gaits for:

1) walking on ice/suspected ice (and no, it's not just walking "slow")
2) walking quickly (yes, this isn't just a faster version of #6; it's an independently created gait)
3) walking silently (there are also subversions depending on the surface - grass, leaves, water, etc).
4) hiking on uneven surfaces
5) walking blind
6) walking while doing none of the above.

While #6 may be my "normal gait", there's no reason why I can't switch to one of the others, even for long durations (and I do for reasons that have nothing to do with hiding from theoretical gait-watching satellites).

dmcSeptember 10, 2008 10:17 AM

I think folks are missing the point of this.

I don't imagine there's any interest at all in this sort of technology for monitoring large numbers of people.

I would guess that the ones who would be most interested in this sort of thing would be the CIA. Remember when they were worrying about telling Saddam apart from his doubles in videos of public appearances, based on ear-lobe analysis? It's a very specific application, with plenty of positive examples, and all that is necessary is to determine whether a new sample matches the known positives. I imagine that the current application would be to analyze lots of footage of Bin Laden, along with surveillance footage of some mountains in Afghanistan, and hope you catch him taking his morning jog.

Of course a target who was aware they were being gait-monitored could probably easily defeat this technology. Just as Saddam could have defeated the ear-lobe analysis by wearing a bandage over his ears. But what are the chances that a target would think they were being monitored for something as arcane as this? And what are the chances that they'll be consistent about trying to defeat it?

RoxanneSeptember 10, 2008 10:24 AM

I predict that parasols and umbrellas will become trendy in areas that don't have frequent cloud cover. This is another reason to move somewhere that it rains a lot. :-)

Davi OttenheimerSeptember 10, 2008 2:13 PM

@ dmc

Correct. This is about finding wanted individuals and sending a rocket from far away to say hello.

@ SteveJ

"...a point when some nation decides that another nation has too many spy satellites pointed at it, and starts shooting them down"

This is less about state sparring and more about a state trying to "surgically" eliminate opponents as they move about.

I am discovering that it is supported most by people who believe that shooting far away is less risky to themselves, because they do not factor in the cost from blowback and environmental harm (or perhaps don't care).

mooSeptember 10, 2008 2:52 PM

Shooting down satellites doesn't seem like a bright idea, but I've been a fan for a long time, of targeted assassination as a tool of foreign policy.

It would surely be more effective, and cause less collateral damage, than the ham-handed approaches the U.S. has been using lately.

bobSeptember 10, 2008 2:57 PM

Great novella. Reminded me of "The Cuckoo's Egg" except in this one the government are active participants instead of neutral parties with their hands in their pockets passively watching their networks be hacked while a single private citizen does their job. Cliff Stoll should be thankful he did his work pre-Patriot Act.

fanfanSeptember 12, 2008 5:27 AM

The article is about shadows analysis and how it can facilitate above-the-head biometrics. To what degree gait biometrics is valid is a different story - and not the contribution of the idea./article. As for walking in the shadows only, I find it hard to believe one would never cross a street where the sun can see you - a few steps should be enough. As for night walks - both city lighting and moonlight may do it for you. Chose to walk in total dark, and you could sneak in...
For those thinking of wearing small rocks in the shoes for a lifetime - i wish them happiness. Same to those making umbrella sales. As for those who worry about the satellites - the paper clearly mentions UAVs - satellites as a possible future option - not soon, UAVs much cheaper. Finally, about altering ones gait...sure you could drag your one leg. But it is no clear evidence to what extend superficial measures - shoes change, etc, may actually affect specific aspects of your signature - it may appear different to you, and still include a specific aspect/spectral pair that makes your gait yours rather than joe's. Yes, there is a probability of error - etc, etc. and unlikely to be more than ~60% correct classification - but that does not need to be actionable intel. It may be one extra piece of info for fusion. Seems like a nice idea to me....

JoeSeptember 16, 2008 12:36 AM

Osama bin Laden has been spotted in the mountains between Afghanistan and Pakistan sporting a big umbrella lately...

Check out his gait under the umbrella, hahahahahahahahahahahahahahah...

What a waste of time/money and intelligence..

/// Joe

Leave a comment

Allowed HTML: <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre>

Photo of Bruce Schneier by Per Ervland.

Schneier on Security is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc..