[huml-comment] Fwd: gait characterization research Re: walk

[Rex Brooks <rexb@starbourne.com>]

I thought I would pass this tidbit along.

Ciao,
Rex

>Date: Sun, 02 Feb 2003 19:54:08 -0700
>From: "cognite@zianet.com" <cognite@zianet.com>
>X-Accept-Language: en-us, en
>To: "L. Reafsnyder" <budoka42@mindspring.com>,  rexb@starbourne.com
>Subject: gait characterization research Re: walk
>X-Rcpt-To: <rexb@starbourne.com>
>X-DPOP: Version number supressed
>Status: U
>
>some tracking down of this research might be relevant to the HuML 
>work. The researchers involved aren't named in this, but it looks 
>related to the Native Dancer material perhaps.
>SC
>
>received:
>
>>Saturday, 1 February, 2003, 09:24 GMT
>>Walk offers clues to identity
>>   Everyone walks in their distinctive way
>>  A new technique of personal identification
>>  is being developed that uses something as simple as the way you 
>>walk to work out who you are, as BBC World ClickOnline's Andrew 
>>Webb reports.
>>   Each and every one of us has our own distinctive way of walking, 
>>called our gait. Many of us share general characteristics, but 
>>physical differences make a difference in the way we move.
>>  Because they have wider pelvises, women sway their hips more than men.
>>  And people from one country can have a different gait to those in another.
>>  What is of particular interest to scientists is trying to make a 
>>computer do what we all do instinctively - know within a split 
>>second that the lolloping or swaggering figure coming towards us is 
>>a close relative, a complete stranger, or one of our best friends.
>>  Researchers at Georgia Tech in the US city of Atlanta are trying 
>>to teach computers to record the precise way we move.
>>  Checking identities
>>  Volunteers walk through a studio while wearing metallic disks. A 
>>computer monitors the position of each disk and builds up a digital 
>>picture of the person's gait.
>>   Person's walk broken down into a computer image
>>When the computer analyses film shot in a different location, it 
>>identifies the volunteer as being the same person.
>>  "Right now we're at the stage of verification," said Professor 
>>Aaron Bobick, associate professor of computational vision at 
>>Georgia Tech.
>>  "If somebody comes in with an ID that says that's John, and then I 
>>see this person walking around outside can the computer tell that 
>>it really is John? The current performance levels are getting us 
>>close to being able to do that pretty well."
>>  The security industry is interested in the technology as a way of 
>>picking a criminal out in a crowd.
>>  The real challenge for scientists is to be able to recognise 
>>someone's gait out in the real world.
>>  Lines of symmetry
>>  At the University of Southampton in England, gait aficionados 
>>believe they are cracking that problem.
>>  So we put them to the test. The first step was to generate a 
>>computer image of me walking through their lab.
>>  Mark Dixon of Southampton University explained: "We film as you 
>>walk and then we bung the film into computers.
>>  "These then pull out a series of numbers, and each time you walk 
>>past the camera you should get the same set of numbers.
>>  "A different person walks past and they produce a different set of 
>>numbers. We then recognise you by that set of numbers," he said.
>>  Each person's walk is broken down into a computer image marking 
>>out lines of symmetry as the body moves.
>>  Gait monitoring
>>  The technique has only become possible in recent years as the 
>>power of microprocessors has accelerated.
>>  The question is whether the 7,000 gigabyte memory bank can tell me 
>>apart from everyone else it has stored.
>>  In a bustling street the computer has difficulty separating me 
>>from all the other movement, but when I am the only person in shot 
>>there is no problem.
>>  Another way to confuse the computer is to radically alter my walk. 
>>The lines of symmetry generated are so different to my normal walk 
>>that the computer is unable to recognise me.
>>  If gait technology ever catches on as a security weapon could it 
>>be that criminals start looking for inventive ways to avoid being 
>>recognised?
>>  Perhaps adopting someone else's walk will become the ideal way of 
>>foxing tomorrow's gait monitoring police.
>>  http://news.bbc.co.uk/2/hi/technology/2712995.stm


-- 
Rex Brooks
Starbourne Communications Design
1361-A Addison, Berkeley, CA 94702 *510-849-2309
http://www.starbourne.com * rexb@starbourne.com