Stunt Woman Tests Apple Watch With Violent Fake Falls (hothardware.com) 96
It seems like everyone's curious about how the Apple Watch 4 detects falls. The Washington Post reports:
In the interest of science, I've tried jumping off ledges and throwing myself onto furniture. The thing never went off. (The feature is on by default only for people older than 65, but I turned mine on.) It's possible, even likely, that the Watch could tell I was faking.
What's important is actual falls, not stunts. Apple says it studied the falls of 2,500 people of varying ages. Yet the company hasn't said how often it catches real falls or sets off false alarms. This isn't like claiming the "best camera ever" on a smartphone -- if Apple wants us to think of its products as life aids, it ought to show us the data. Even better: peer-reviewed studies. Apple's disclaimer says: "Apple Watch cannot detect all falls. The more physically active you are, the more likely you are to trigger Fall Detection due to high impact activity that can appear to be a fall."
But there's now also a new video by the Wall Street Journal that tests the watch's fall-detecting capabilities with a professional stuntwoman. Hot Hardware reports: The Wall Street Journal found that the Apple Watch did a very good job of detecting a serious fall while ignoring insignificant or outright fake falls. The stunt double performed a series of falls that are similar to falls in the slides that Apple showed in its keynote explaining the feature. In the testing, the watch was able to identify those falls and offer to call emergency services.
The most interesting part is that even though the stunt woman pulled some serious fake falls, complete with Hollywood-style tumbling down a hill, the Apple Watch was able to figure out if the fall was fake and didn't offer to call emergency services.
The Journal's reporter credits the watch's gyroscope and accelerometer, which can monitor numerous factors including both speed and wrist trajectory. Their conclusion?
"Turns out the Apple Watch really does know when you're just playing around."
What's important is actual falls, not stunts. Apple says it studied the falls of 2,500 people of varying ages. Yet the company hasn't said how often it catches real falls or sets off false alarms. This isn't like claiming the "best camera ever" on a smartphone -- if Apple wants us to think of its products as life aids, it ought to show us the data. Even better: peer-reviewed studies. Apple's disclaimer says: "Apple Watch cannot detect all falls. The more physically active you are, the more likely you are to trigger Fall Detection due to high impact activity that can appear to be a fall."
But there's now also a new video by the Wall Street Journal that tests the watch's fall-detecting capabilities with a professional stuntwoman. Hot Hardware reports: The Wall Street Journal found that the Apple Watch did a very good job of detecting a serious fall while ignoring insignificant or outright fake falls. The stunt double performed a series of falls that are similar to falls in the slides that Apple showed in its keynote explaining the feature. In the testing, the watch was able to identify those falls and offer to call emergency services.
The most interesting part is that even though the stunt woman pulled some serious fake falls, complete with Hollywood-style tumbling down a hill, the Apple Watch was able to figure out if the fall was fake and didn't offer to call emergency services.
The Journal's reporter credits the watch's gyroscope and accelerometer, which can monitor numerous factors including both speed and wrist trajectory. Their conclusion?
"Turns out the Apple Watch really does know when you're just playing around."
It probably measures (Score:4, Interesting)
It probably can tell real falls from fake ones by measuring rapid decelleration. In a real fall you would usually come to a stopped state almost instantaneously while in a fake fall you would absorb some of the impact by cushioning the impact (slowing the decelleration) with your arms and legs. If this is right then this watch wouldn't detect someone falling off a roof and getting impaled on fence.
Re:It probably measures (Score:5, Funny)
If this is right then this watch wouldn't detect someone falling off a roof and getting impaled on fence.
Clearly, studies are needed.
Re: It probably measures (Score:1)
So for this test we need some die hard Apple fans?
Re: (Score:3)
You forgot to add that the studies must be peer reviewed and public.*
* So Samsung all the other Androids can profit from the research Apple does and not have to duplicate it (even if it's tongue in cheek studies like the ones you suggested).
Re: (Score:2)
In completely unrelated news, Foxconn has been replacing the suicide nets around their factories with spikey fences and high-speed video cameras...
It's not just that (Score:2)
It probably can tell real falls from fake ones by measuring rapid decelleration.
That is probably part of it (if you didn't land hard enough, why signal for help?) but there's at least one other aspect - in the keynote Apple showed how they also measured the wind-up if you will - the arm motions that preceded a fan, especially something like tripping or falling where the arms have pretty detectable and nearly involuntary actions that you can use as an indicator along with how hard you land. The involuntary
Re: (Score:1)
Hope you didn't fall unconscious/faint if your arms have to do something to trigger it.
Re: (Score:2)
I think it is more likely to be the opposite. If sharp motion is detected before the fall, it most likely means that your are bracing yourself, minimizing the damage even with a hard fall. It happens naturally.
If you don't, it means you may have fainted or that you are too weak to control your fall. And that's the worst case.
Re: (Score:1)
I think it is more likely to be the opposite. If sharp motion is detected before the fall, it most likely means that your are bracing yourself, minimizing the damage even with a hard fall. It happens naturally.
If you don't, it means you may have fainted or that you are too weak to control your fall. And that's the worst case.
Bracing for the fall is actually one of the worse things you can do. Stuntmen, martial artists, etc. who practice falling learn to not brace but rather "roll with' the fall which is why they can jump off a 20 foot building without injury, where a regular person can brain themselves just falling over.
Re: (Score:2)
Bracing for the fall is actually one of the worse things you can do.
True.
But without specific training/practice, it is a nearly involuntary action.
Fainting is just another example (Score:2)
Fainting would have easily detectable arm motion also because your arms would move in a very rag-doll like manner without control until you hit the ground. Just like tripping and slipping have distinguishable precursors of movement to detect and use as a judge of if it should prompt to alarm or not.
Re: (Score:2)
You missed an article before "other", Ivan. And it should be "putting it in context".
P.S. Sentences don't end themselves.
Re: (Score:2)
cvrses! foiled again by moose and sqvirrel.
Re: (Score:2)
sorry gimmick
Sounds like someone's jealous...
Re: (Score:1)
Nobody's claiming perfection. That would be silly. But lives will be saved. In what way is this a "sorry" gimmick?
Re: (Score:3)
I wonder if its also having a look at your bodies behavior too. A stunt person who pulls off a successful stunt probably is going to exhibit healthy vital signs, albeit with a bit of extra andrenalin. But someone whos just injured themselves going to have a significant change in their bodies vital signs , heart rate etc as the body does whatever the body does whenever its compromised. And I'd wager that a highly unfit person, or someone who's elderly will have a more marked change in vital signs than really
Re: (Score:3)
I wonder if its also having a look at your bodies behavior too. A stunt person who pulls off a successful stunt probably is going to exhibit healthy vital signs, albeit with a bit of extra andrenalin.
I think the biggest difference is that the stunt woman has trained to do falls that _look_ spectacularly dangerous _without_ actually hurting herself. If she got a fall wrong that would be much more likely to be detected.
Re: (Score:2)
Yep, that's the combination I used when I came up with the idea in a story (see other post). Gee loads and physiological changes. If it could monitor blood-pressure too (tricky without more invasive sensors) it could detect injuries not accompanied by massive forces, like a serious cut or stab that's bleeding out (for the guys above worried about falling onto a sharp fence post. ;)
boring (Score:4, Insightful)
Look it's going to detect some real false and it's going to have a false positive rate and a missed-true rate. no matter what. so what ecaxly are we discussing here. Joanna Stern at the WSJ dod the exact same thematic test with a stunt person and came to the exact oppostie conclusion.
Bottom line is a senior citizen who lives alone is better off with this than without it. Anyone disagree?
Re: (Score:2)
I disagree. False positives will lead to a 'boy who cried wolf' scenario for when people actually fall over.
Re: (Score:1, Informative)
I turned the feature off after several false alarms in less than a day of normal use
Re: (Score:2)
I turned the feature off after several false alarms in less than a day of normal use
Liar.
Yeah but can it detect... (Score:3)
Niagra Falls?
Slowly I turned...
Step By Step...
Inch By Inch...
Re: (Score:2)
The "you should feel" meme is old and you should kill yourself.
Re: (Score:2)
The "do you need a hug" meme is ancient and also sexual harassment.
(I do)
Re: (Score:2)
Your Apple Watch is sentient (Score:2)
It's possible, even likely, that the Watch could tell I was faking.
If your Apple Watch can detect when you fake fall versus when you real fall, then it is sentient. In the future, you will have to ask it nicely to do anything which may result in harm or damage to your Apple Watch. Please watch this documentary [wikipedia.org] for additional details.
Re: (Score:3)
1. An Apple Watch may not injure a human being or, through inaction, allow a human being to come to harm.
2. An Apple Watch must obey the orders given it by human beings except where such orders would conflict with the First Law.
3. An Apple Watch must protect its own existence as long as such protection does not conflict with the First or Second Laws.
Yeah Right... (Score:2)
"The most interesting part is that even though the stunt woman pulled some serious fake falls, complete with Hollywood-style tumbling down a hill, the Apple Watch was able to figure out if the fall was fake and didn't offer to call emergency services."
Yeah right. How do we know the watched was that smart and just didn't trigger properly on what should have been perceived as a massive fall. The type of stunts stunt people do are not normal activities that should be ignored for the rest of us.
Re: (Score:2)
The type of stunts stunt people do are not normal activities that should be ignored for the rest of us.
"You appear to be experiencing domestic abuse. Would you like me to call emergency services?"
Eh, it was just an advertisement. (Score:5, Interesting)
That video was just an advertisement, it praised all the features, including: "the battery life has been solid, lasting a day and a half" !
And it claims this is a good investment for older people who need fall detection, just because the $400 device did detect a falls they tried. Will the older people remember to charge it every night? How will it detect their falls when they go to the toilet during the night charging time? At the same time there are specialized fall detectors for the elderly, how about doing a comparison?
Many Apple slashvertisements lately...
Re: (Score:2)
"A day and a half".
Is this a part of the new "all day battery" catchphrase, as if it were some innovative new feature?
I'll stick with my four day Gear, thank you.
Re: (Score:2)
And it claims this is a good investment for older people
My grandma has a device like this, it cost her insurance company 70EUR on her behalf. $400 for a toy gadget that needs an expensive phone with it is not "a good investment".
Apple Testing (Score:3)
How can I get a job tipping over old people?
Re: (Score:2)
Become a police officer.
Methodology failure? (Score:1)
I wonder if her methodology also included simulating what happens after a fall resulting in injury (tried to RTFA but blocked due to ad blocker detection), such as lack of movement, labored movement, erratic heart beat, etc. It might be that the watch looks for two or more conditions before triggering an alert, for example a high G event followed by slow twisting movements or no movement at all.
'serious fall' criteria: (Score:2)
If your heart is still beating it is not a serious fall.
Re: (Score:2)
Or better yet - what they actually care about is fall->unconscious. If you are still conscious, you don't need your watch to call 911, you can call yourself.
After a real fall, you don't get up. You make no attempt to minimize the fall damage, etc.
Re: (Score:3, Informative)
My mother broke her hip a few years ago by tripping over the edge of a chair and landing on tile, trying to get to her phone on the kitchen counter. She would've had to crawl, with a broken hip and various smaller injuries, some forty feet to the kitchen, remember exactly where her phone was, and somehow reach the top of the counter. Conscious, but in severe pain and unable to reach her phone. (And what actually came in handy was shouting at Alexa to call me.)
Re: 'serious fall' criteria: (Score:1)
The watch does call automatically if no motion is detected.
Makes sense given that the falls were faked (Score:3)
The whole point of a stunt fall is to look like a real fall but not produce the kinds of impacts that'd result in injury. It makes sense that trying to fake a fall that way would result in the watch deciding it wasn't a real fall. There wouldn't have been, for instance, the kind of sharp, sudden impact you see in a real fall because the stuntperson would be spreading the impact out over time so they wouldn't break bones.
Alternative explanation... (Score:3)
I somehow think it is more likely that it's not very good at recognizing serious falls in the first place.
Unless the stunt woman actually went and caused real serious injury to herself on at least one fall and it *did* correctly identify that as a real fall, then I'm more inclined to think that this is feature that doesn't work as well as they might advertise it to.
Re: (Score:2)
"I somehow think it is more likely that it's not very good at recognizing serious falls in the first place."
Of course. Interesting that I had to read through almost all of the comments to get to one that points out that likely answer. -- the device, like a lot of digital stuff, just doesn't work very well. How to test?. Strap some of these things on individuals in sports where participants legitimately fall down a lot.-- American/Canadian football for example. Or put them on folks just learning to ski o
Re: (Score:2)
"I somehow think it is more likely that it's not very good at recognizing serious falls in the first place."
Of course. Interesting that I had to read through almost all of the comments to get to one that points out that likely answer. -- the device, like a lot of digital stuff, just doesn't work very well. How to test?. Strap some of these things on individuals in sports where participants legitimately fall down a lot.-- American/Canadian football for example. Or put them on folks just learning to ski or ice skate.
No matter how good or bad it is, I am relatively sure that the Apple Watch's fall-detection is MUCH better than:
1. "Lifeline"-type Panic-button Pendants. Those are made for about $2, and simply don't have any real computational power.
2. Nothing. Which is what most people have...
Re: (Score:2)
This. You need to know all four squares of the grid.
These people prayed and got better! But how many prayed and didn't get better, how many got better without praying...
She needs to go to a party (Score:1, Flamebait)
Re: (Score:1)
What went wrong? (Score:3, Funny)
I don't know why I'm seeing this story, since I have Ad-Block turned on.
Re: (Score:1)
Ad-Block allows some "acceptable" ads through if they paid-off the developers.
Switch to uBlock Origin.
Oh the Humanity! (Score:1)
Crash-Test-Dummy: "Dammit, filthy humans are taking our jobs!"
Apple (Score:1)
You're falling wrong.
Re: (Score:2)
It sounds like it could detect the falls actually portrayed by Apple when recreated. Is it actually just detecting those motions and everything else is a "fake" fall? 2500 isn't very many, and it is far more credible that it is easily fooled and missing actual falls than that it is miraculous as not being tricked by fake falls.
You forgot to say "In my not so humble opinion".
As in martial arts ... (Score:5, Informative)
As in martial arts, stunt people learn how to go to the ground while minimizing the impact on their body.
So a fake fall is quite different from a natural fall, even if it looks for the audience the same (or similar).
Proven! (Score:2)
Well, that proves it. Time to put it on over a million people's wrists without worrying about any false callouts.
Re: (Score:2)
Well, that proves it. Time to put it on over a million people's wrists without worrying about any false callouts.
The likelihood of significantly high numbers of truly false "Callouts" is drastically reduced by the fact that the Watch counts-down audibly and visibly before automatically dialing 911 (and your emergency contact(s)). The (probably good) assumption is that if you are aware-enough to cancel that "self-destruct" countdown, you can probably also fairly accurately assess your need for an emergency response. But if you DON'T respond, it is also damned likely that you CAN'T respond...
Mine seems to work well (Score:5, Interesting)
Iâ(TM)ve had mine for a couple of weeks. Iâ(TM)ve had one false fall detection happen when I was using my fist to literally hammer together an IKEA style bookshelf. The high G impact was enough. Iâ(TM)ve had one real detection when I went down hard riding my mountain bike. In both cases it alerted and asked if I needed help. BTW: Iâ(TM)m seeing two solid days of battery life even when using GPS and heart rate monitoring for multiple hour rides.
Sounds to me like neural network programming (Score:1)
Sounds like they trained a neural network on a lot of falls.
Which means, they don't really know what they actually trained it to do either.
Re: (Score:2)
Sounds like they trained a neural network on a lot of falls.
Which means, they don't really know what they actually trained it to do either.
Just because that's the way YOU would approach the problem, is not any assurance that Apple took JUST that approach.
Re: (Score:1)
Well your answer sounds like the kind of answer YOU would give, but that's not any assurance that an Apple or I would answer my post that way.
Re: (Score:1)
Well your answer sounds like the kind of answer YOU would give, but that's not any assurance that an Apple or I would answer my post that way.
You're right.
Frankly, Apple probably wouldn't have answered at all, given the sheer volume of mail they get every day.
I "invented" this a year ago. (Score:2)
Over a year ago (don't recall exactly when, the book was published last December) I used a similar device in The Eridani Convergence when the protagonist hurls himself from a moving car (autonomous, and it had been hacked to kidnap him) in the middle of nowhere. He wakes up in a hospital room:
Re: I "invented" this a year ago. (Score:2)
You should have patented the idea (*), then Apple would owe you licensing fees! Or they would perceive you as a threat and sue you out of existence. Maybe that's more likely.
(*) Technically one does not patent a concept or idea. But as someone who creates real complex software system that millions of people people have used, when I look at a software patent I see an idea, not an implementation.
Re: (Score:2)
gee loads? You said a car, not a horse.
Re: (Score:2)
Funny man. ;)
I tend to write dialog phonetically. "The g loads" almost looks like a typo, and "the G loads" looks like I'm referring to the gravitational constant. ;)
Or maybe the character was yelling "Gee whiz!" very loudly as he jumped....
Re: (Score:2)
Over a year ago (don't recall exactly when, the book was published last December) I used a similar device in The Eridani Convergence when the protagonist hurls himself from a moving car (autonomous, and it had been hacked to kidnap him) in the middle of nowhere. He wakes up in a hospital room:
I like to think that someone at Apple read that and thought it was a good idea, but probably they were already working on it. Or they worked fast. ;)
I can beat that by about 5 years.
In fact, when the iPhone first started having 3-axis Accelerometers, I considered making an "Emergency Call" App with built-in Fall-Detection.
Unfortunately, at the time (and maybe still), Apple prohibited that type of App from being published in the App Store.
But in this case, this instrinsic feature of WatchOS 5 is a obviously a LOT deeper-integration than a simple App can attain, even without the App Store restrictions. So it's an entirely different case.
BS (Score:3)
Okay, this is a bunch of bullshit.
My wife has an Apple Watch. Its detection algorithms are extremely inconsistent: it frequently doesn't detect that she's exercising or her heart rate. It frequently doesn't detect that she has raised her wrist. Etc.
The actual title of this article should be: "Apple Watch Cannot Reliably Detect Falls." Because that's the far simpler explanation: not that it has some fancy algorithmically-generated profiles for "real falls" vs. "fake falls," but that it has one profile for "falls" that is unreliable.
Re: (Score:2)
Okay, this is a bunch of bullshit.
My wife has an Apple Watch. Its detection algorithms are extremely inconsistent: it frequently doesn't detect that she's exercising or her heart rate. It frequently doesn't detect that she has raised her wrist. Etc.
The actual title of this article should be: "Apple Watch Cannot Reliably Detect Falls." Because that's the far simpler explanation: not that it has some fancy algorithmically-generated profiles for "real falls" vs. "fake falls," but that it has one profile for "falls" that is unreliable.
You didn't say which Series your wife's Apple Watch is.
I've fallen and I can't get up! (Score:3)
Our insurance company sent my wife one of those devices at no cost following her heart mitral valve replacement and hip replacement. It turned out to be touchy and temperamental. She'd just set it down on the coffee table and it would trigger dial the company. The service person had to go through a script before they'd hang up. While hanging around her neck it would swing back and forth and eventually bang against her sternum, setting it off. She averaged 3 to 5 false triggers per day. On the only day she could have used it we were walking to the stadium to watch our grandson play baseball. We went single file through a gap in the curb and I heard a thud behind me. She had fallen face first flat onto the concrete. The device didn't trigger. I drove her to the doctor and she got a 1.5" cut sewed up. Her face was black and blue for months because she was on blood thinner. We boxed the device up and sent it back. Fortunately, it didn't cost us anything.
I suspect that the Series 4 Apple Watch would behave the same way.
Re: (Score:2)
Our insurance company sent my wife one of those devices at no cost following her heart mitral valve replacement and hip replacement. It turned out to be touchy and temperamental. She'd just set it down on the coffee table and it would trigger dial the company. The service person had to go through a script before they'd hang up. While hanging around her neck it would swing back and forth and eventually bang against her sternum, setting it off. She averaged 3 to 5 false triggers per day. On the only day she could have used it we were walking to the stadium to watch our grandson play baseball. We went single file through a gap in the curb and I heard a thud behind me. She had fallen face first flat onto the concrete. The device didn't trigger. I drove her to the doctor and she got a 1.5" cut sewed up. Her face was black and blue for months because she was on blood thinner. We boxed the device up and sent it back. Fortunately, it didn't cost us anything.
I suspect that the Series 4 Apple Watch would behave the same way.
It's a pretty big assumption that a freebie pendant that probably cost $2 to make in China would have the same computational power to determine actual falls as the Software and Hardware inside the Apple Watch.
Emergency False Alarms (Score:2)
When I saw this feature my 1st thought was "I wonder how the emergency services call centres will deal with a higher volume of false calls?"
This isn't the first (v)blogger to try out this new Apple watch feature and, from what I've seen so far, it looks like it's takes more than a theatrical fall to trigger it.
It takes no great technical genius to set up something that starts a 1 minute timer after the accelerometer detects an impact and call 911/999/111/etc
What Apple's not telling is how they've mitigated,
Re: (Score:2)
es, I know the watch asks you if you are ok before calling but you might be distracted long enough to not notice or be able to respond.
If you are not responding to an audible countdown and message then there is likely something wrong, and it SHOULD call.
as always - FA (Score:2)
Re: (Score:2)
I have to tell my watch how fucking old I am now? What else does it need to know?
No, Hater.
When you sign up for an AppleID, IIRC, DOB is part of the required information. I would assume it uses that.