Facial-recognition tech now used to greet hotel guests
'Welcome back to the Grand, your Highness (82±13%)'
Facial recognition technology has long been a holy grail for security agencies, and a bogeyman for privacy lovers and libertarians. One company, however, says that reasonably priced face-recognition tech has uses outside the security field: in particular, in the service sector.
At the Counter Terror Expo this week in London, Derbyshire firm Zycomm was showing off its wares to the assembled cops, spies, security types and mercenaries. Zycomm mainly distributes two-way radios - for security guards, for instance. But they also offer a facial-recognition system for use with security cameras, aimed at customers such as shopping centres.
"This stuff used to be strictly government-agency business," Zycomm engineer Steve Ball told The Reg.
"But now you can get a useful system like this in the £20,000 bracket - it's within the reach of commercial customers."
Zycomm's setup can zero in on faces from any camera, but Ball said that like most face-spotter software it only starts to show useful levels of performance where people need to move through a chokepoint facing the camera - ideally looking up into it and well lit.
"Doors are OK," says Ball. "Escalators are good - we all look up when we're on an escalator."
The biz-grade spotter has an 80 or 90 per cent chance of recognising someone in its files in such circumstances, according to Ball, provided the file image is a good one. Zycomm's special trick is that the system can then send a text message to a portable radio, saying who it thinks it has identified, where, and with what level of confidence.
Typically this might be used to notify patrolling security guards that a known shoplifter had entered their complex. In this case the file images would be those of convicted crims routinely circulated by police - Ball doesn't recommend that his clients start treating people as thieves without solid proof.
In theory, eagle-eyed CCTV operators could do this without automated assistance, but very few commercial security teams can afford enough staff to individually monitor every camera. Typically there will be only one person watching all the cameras in a building, and this person will also be manning a reception desk, answering calls, issuing keys etc. So an educated guess by the automatic face-spotter is better than nothing.
But Ball says it doesn't have to be all about security. He says at least one hotel has bought the system; but instead of having face-files of thieves, confidence tricksters, tarts etc the system is set up to recognise returning customers. Thus the doorman or concierge can receive a message on their handset to the effect "His Highness the Maharajah of Grogpore, Grade 1 high roller, approaching main door, 82 per cent confidence", and so greet valued clients by name.
It is, of course, every free-born Briton's right to take pictures of people and keep them - particularly on his or her own property - so there isn't any legal problem with keeping private face files provided the Data Protection Act is complied with. Nonetheless the wise hotelier would make sure that his customers were happy to have pictures added to all the other information a company likes to record about its clients.
As for whether it works, one might say that just as 80-90 per cent accuracy is never going to sustain the man-tracker panopticon aspired to by the Home Office, neither is it good enough in the case of handling important customers. But that, literally in this case, would depend on your viewpoint. ®
Sponsored: Beyond the Data Frontier