In Recognition of Recognition

I was reading New Scientist web site, and looking at the top 10 news stories of the year, and I noticed that one highlighted the rise of recognition software.

If you are lost in a big city, or are late for a meeting, and you don't know where you are, you can take a picture with your mobile phone, and text it to a remote server. Software will then attempt to match the picture with images in their database, and inform you where you are. Spooky, huh? Click here to see how it's done. It can place you to within 1 metre of where you are standing, which is better than GPS systems, which can only resolve position down to 10 metres currently.

Recognition is largely down to matching patterns with those that you have seen before. Since computers are very good at analysing patterns (as in the case of vector recognition, above), and using mathematical analyses, there has been more and more of this technology in abundance. Another story on the same site describes how a piece of software can automatically caption your photos, by listening to the way people discuss them. HP have developed software which records you and your friends talking about your photos, analyses them on a computer, then produces captions for the photos. Over time, the details are the things that are missing from photos - memories fade, and the names of individuals and places get hazy. This software can help permamently describe a picture.

It's true though - if you hand your holiday photos to a friend, unless you are around to describe each picture, the shot usually has no meaning. It is those descriptions that bring the image alive, not just the contents of the shot, but the events that happened before and after it was taken.

Music has not escaped the realms of pattern recognition either. Ever heard a tune you liked, but just never knew who recorded it, or just plain forgot? In the UK, you can contact Shazam Entertainment on 2580, hold your phone close to the source of the music for 30 seconds, and a remote server texts you the name of the track with the artist. Pretty cool.

And finally, OCR (Optical Character Recognition) is not new, but Sun have developed a handwriting recognition software, which apart from being a cool toy, could also assist forensic analysis of handwriting in criminal investigations.