Handwriting is a skill that varies person to person. Everyone has their own version of handwriting, someone’s is pretty beautiful and someone’s is a messy scrawl. Sometimes our own handwriting fluctuates from one line to the next. So it’s nearly impossible for a computer to replicate them. However, with the development of new software, researchers at University College London (UCL) are getting close to teaching the computer to emulate anyone’s handwriting.
Known as “My Text in Your Handwriting,” this new algorithm is able to replicate your handwriting to an impressive degree by scanning what you’ve written on a piece of paper. It can also produce any word you wish in your handwriting.
To capture your handwriting or whatever messy scribble you call it, you will have to write on four A4-sized papers. The software will scan the text and convert it into a thin, skeletal line. Computers along with a human moderator then break down the line, assigning letters and their position within a word. Also, they will look for “splits,” where the line converts to a “ligature” from letter. “Ligature” is the extra bits needed for joined up penmanship. Finally, there will be links, indicating that two individual marks are actually the part of the same letter.
The software then references and adapts your previously scanned example to imitate your handwriting style. From those examples, it looks for the one that best works for the word it’s trying to create. It ensures that the same combinations and letters aren’t used more than one time by applying a degree of randomness.
After selecting one of your examples, it will consider the appropriate spacing between each letter, as well as the height of each letter and where it stands in the line. Finally, the software adds the “ligature” to the computer generated piece, adding some basic texturing to copy the quality and color of the ink.
The results are pretty amazing. To test its effectiveness a group of people were asked to distinguish between handwritten envelopes and the ones created by that software. People get puzzled by those envelopes and chose incorrectly 40 percent of the time.
Co-author, Dr. Oisin Mac Aodha (UCL Computer Science), said: “Up until now, the only way to produce computer-generated text that resembles a specific person’s handwriting would be to use a relevant font. The problem with such fonts is that it is often clear that the text has not been penned by hand, which loses the character and personal touch of a handwritten piece of text. What we’ve developed removes this problem and so could be used in a wide variety of commercial and personal circumstances.”
This computer generated message can be used to send people greetings, where handwritten messages are more appreciated than something typed on a paper. Another concern is it may be used by the criminals to copy signatures, but the researchers say close examination with a microscope can reveal whether it’s handwritten or machine generated.