Paul Gilster, Correspondent
We've learned a lot about computers in the past 25 years, but we're entering a new era. Now it's time for computers to learn about us.
Tasks that we routinely perform with menus and mouse clicks are becoming the object of voice commands. GPS units scold us audibly when we take the wrong turn. Imminent breakthroughs in face recognition will make it possible for machines to identify us in a crowd, with implications for security and personal privacy.
Many of these changes are welcome and solve genuine problems.
Charlotte-based Yap (
www.yapme.com) says that cell phone companies admit in meetings that two-thirds of their teenage customers have sent or read text messages while behind the wheel.
If the idea of an inexperienced driver approaching you in an SUV while thumb-typing text into a cell phone doesn't scare you, nothing on the planet will.
Enter speech recognition, which Yap hopes to turn to the public good even as it creates a lucrative business market. Sure, voice recognition has been available for years as part of programs such Dragon NaturallySpeaking, and it has become a supple and effective tool.
But the cell phone market is where the breakout will occur. Think of the billions of cell phones saturating the market and the potential for hand-held software that lets you manipulate your phone in the car without having to glance away from the road ahead of you.
What Yap is working on is a way to send text messages by simply speaking, but it translates into more useful services such as searching the Net by voice, checking the itinerary for a trip or connecting to instant messaging services by speaking.
Yap does this by capturing what you say in a program loaded on your phone, uploading your words to its servers for translation. Off goes the text message or whatever traffic you intend, from querying Web services such as Amazon to searching entries on Wikipedia.
An even better solution for the cell phone problem in cars is a legal ban on using them there. But until we pass such legislation, the more technology can do to help, the better.
Besides, companies such as Yap and Massachusetts-based Vlingo (
www.vlingo.com), which specializes in voice recognition on phones, aren't just about driving. With cell phones everywhere, services we would like to use face the challenge of tiny screens and absurdly ineffective keyboards. Voice is the obvious solution for mobile computing's next phase.
With speech recognition a growing market, it's no surprise that another intractable problem seems on the way to a solution. The Net has become a vast storehouse of images, abetted by the wild popularity of photo-sharing services such as Flickr. The problem with images, though, is finding them.
Most search functions on the Web find what you're looking for by searching text that is associated with the image. The choice of that text depends upon the skill of the person who added it and the level of detail brought to the task.
But what if we could search photos directly?
Swedish startup Polar Rose (
www.polarrose.com/) has just that in mind. The company offers a free plug-in program for the Firefox Web browser, with an Internet Explorer version on the way. Now in beta testing, the software detects people in public images and identifies them.
The method: Analyzing the picture and converting a face into an indexable model. Digital photos begin to acquire the indexing capacity of text documents, making running an image search a more accurate proposition.
Face recognition, of course, is about far more than indexing a photo collection on Flickr. The security aspects are immense.
Border security could be enhanced with face recognition systems trained to scan for known terrorists, for example, while businesses can acquire an improved model for controlling access to sensitive information. In Glasgow, researchers are working with methods that create a composite face from a cluster of photographs of the same person.
Thus far their algorithms have proven considerably more effective than previous methods, but time will tell whether their results can be confirmed in real-world use. Nonetheless, couple accurate face recognition with local security cameras -- a system that in Britain is becoming ubiquitous -- and you have set up a way for government to track personal movements and identities.
As always, technology reminds us of the need to define privacy and respect its limits.
This is another case where a technology has been developing for years but remains largely unnoticed in the consumer sector. Facial recognition programs of one kind or another have already found a home in Las Vegas, where casinos need to identify known card cheats at their tables.
The trick is moving from a largely static, intensive environment to a moving crowd.
Even with the best cameras, pulling a face out of the flow of people emerging from a New York subway is going to be no easy business, perhaps offering a respite for those of us whose view of the future still includes the face of Big Brother.
Get $150+ in coupons in every Sunday N&O. Click here for convenient home delivery.