MobileRead Forums - View Single Post

cervezas · 10-26-2005, 02:43 AM

I thought for a long time that voice recognition was going to be a killer app for mobile devices, and I'm sure it will still be fairly important. But I've started to wonder if people will really find this a natural and convenient way to interact with their mobile devices. It's not exactly what you'd call discreet, is it?

I've been thinking instead about high resolution cameraphones and how they could be used to recognize text and other symbols for dramatically simplified input to the device. Xerox just announced a breakthrough that purports to turn your mobile phone into a scanner with OCR running on a backend server. Snap a picture of some text and in seconds it comes back to your phone as plain text. With software on the device that can recognize simple text patterns--a URL, someone's contact information, the VIN number of a car, an ISBN number--you could navigate to a web site, or enter a contact, or download car repair records, or order a book from Amazon, all with a single click of the camera button. I see a whole new way for people to interact with any kind of text or coded identifier that they encounter as they go through their day.

Could be very fun, interactive stuff that would take off with young people. Imagine wearing a T-shirt that has your own unique symbol screened on it. Someone snaps a picture with their cameraphone, your symbol is identified on a server somewhere, which initiates a request on your phone for an IM chat with the person who took the picture. Or it takes them to your personal website or blog. Or let's them listen to what's playing on your music player at that moment.

With more intelligent devices, more complex language patterns could be "understood" and some truly amazing but simple applications could emerge. Brian keeps mentioning Jeff Hawkins' work with Numenta (an interest of mine as well) and this area of text interpretation seems like it might be amenable to even some fairly early stage technologies based on HTM.

I've been blogging about this and expect to put some more thoughts on it to page soon if anyone is interested. Cough! Shameless plug.

10-26-2005, 02:43 AM	#3
cervezas palm & java hacker Posts: 52 Karma: 251 Join Date: Oct 2005 Location: Manitou Springs, CO Device: Visor, T3, i500, iQue ...	I thought for a long time that voice recognition was going to be a killer app for mobile devices, and I'm sure it will still be fairly important. But I've started to wonder if people will really find this a natural and convenient way to interact with their mobile devices. It's not exactly what you'd call discreet, is it? I've been thinking instead about high resolution cameraphones and how they could be used to recognize text and other symbols for dramatically simplified input to the device. Xerox just announced a breakthrough that purports to turn your mobile phone into a scanner with OCR running on a backend server. Snap a picture of some text and in seconds it comes back to your phone as plain text. With software on the device that can recognize simple text patterns--a URL, someone's contact information, the VIN number of a car, an ISBN number--you could navigate to a web site, or enter a contact, or download car repair records, or order a book from Amazon, all with a single click of the camera button. I see a whole new way for people to interact with any kind of text or coded identifier that they encounter as they go through their day. Could be very fun, interactive stuff that would take off with young people. Imagine wearing a T-shirt that has your own unique symbol screened on it. Someone snaps a picture with their cameraphone, your symbol is identified on a server somewhere, which initiates a request on your phone for an IM chat with the person who took the picture. Or it takes them to your personal website or blog. Or let's them listen to what's playing on your music player at that moment. With more intelligent devices, more complex language patterns could be "understood" and some truly amazing but simple applications could emerge. Brian keeps mentioning Jeff Hawkins' work with Numenta (an interest of mine as well) and this area of text interpretation seems like it might be amenable to even some fairly early stage technologies based on HTM. I've been blogging about this and expect to put some more thoughts on it to page soon if anyone is interested. Cough! Shameless plug.