Ask any writer, and there’s a good chance he or she will tell you how great it feels to physically write words on a piece of paper. While typing is much faster, and a lot more efficient, something just feels so good when putting pen to paper. In the case of tablets — which tend to have keyboards too wide for dual thumbs, yet too small to type on like a full-sized keyboard — handwriting recognition can be the most efficient way to jot down some words, somewhat satiating that pen-to-paper desire. Unfortunately, handwriting recognition isn’t exactly the peak of refined technology at the moment, and it generally isn’t precise enough to pick up everyone’s wildly different handwriting styles with any real accuracy. Scientists at the Karlsruhe Institute of Technology (KIT) might have found a different solution — a band that detects the motions and gestures of the wrist, and can translate that into writing.
Old and busted
Rather than relying on a fickle screen, KIT’s wristband — or glove, depending on how you view fashion — allows a user to write words in the air, which the system can detect with the help of gyroscopes and accelerometers. The gestures are then sent to a computer over a wireless connection, then the computer makes sure the user was actually writing, rather than, for example, flailing wildly for no reason. So if you are waiving hello or vigorously scratching something, your computer won’t suddenly have a bunch of gibberish written on-screen.
Rather than standard gesture recognition, the airwriting glove employs pattern recognition as well. A statistical model has been created for the characteristics of each letter of the alphabet, taking different writing styles into consideration. For instance, the airwriting system can recognize the difference between writing in all caps (of which some of us are guilty). In total, the recognition system has a bank of 8,000 words, with a general error rate of only 11%. A large chunk of that error rate comes from individual writing styles, and when the system is calibrated for someone’s specific style, the error rate drops to just 3%.
Aside from expanding the system’s wordlist, the KIT team aims to make the device smaller, and thus more comfortable to wear. Further than that, the team would like to integrate the system directly into a smartphone, which would mean you could simply wave your phone around in the air and generate words on its screen.
For their research, the team received a Google Research Award of $81,000 to aid in the development of the system.