The "V" hand sign is a gesture that has made its way to group photos and selfies, and most of us know that this gesture is interchanged as a peace sign or a sign for being kawaii.
But the hand gesture has terrifying associations: it is a victory sign used not only by military troops, but by terrorists as well.
Images in which terrorists flash the "V for victory" sign often show this: a person has his face and head completely covered with a scarf or hood to hide his identity while he stands over the corpse of a victim, flashing the "V" sign.
Because the person's face and head are hidden, law enforcement is prompted to find more effective ways to identify these people, such as voice identification. However, this is not a straightforward or easy method.
'V For Victory'
Now, a group of scientists in Jordan have developed a method to identify terrorists: by distinguishing people in photos from the unique way they make V signs. The method is not new, but the it has not been investigated in this manner until now.
Led by Ahmad Hassanat of Mu'tah University, the team took advantage of the size of the finger and the angle between fingers as a biometric measure, much like the fingerprint. Experts say hand shape differs widely from person to person, if the details can be measured without errors.
Still, the task of detecting a person through a small section of their hand is very difficult. There's also the matter of how much information could be extracted from images that contain the V sign.
The research team first asked 50 men and women of different ages to make a V sign with their right hand. They captured the image against a black background using a camera phone with 8 megapixels. In total, they had produced a database of 500 images.
Combining Methods
Hassanat and his team pointed out that most real-world images have low resolution. This limits the amount of information they could gather, so they decided to limit their analysis to detecting the following: the end points of the two fingers, the lowest point of the valley between the two fingers and the two points in the palm of the hand.
The team was able to analyze various triangle shapes between points, the angles they make and their relative size, among many other aspects. Next, they measured the hand shape through various statistical calculations. When both methods are combined, the team could use 16 features in identification.
Using two-thirds of the images in their database, researchers trained a machine-learning algorithm to recognize different V signs. They tested the algorithm by using the remaining images.
"There is a great potential for this approach to be used for the purpose of identifying terrorists, if the victory sign were the only identifying evidence," the researchers wrote, adding that the algorithm combined with the statistical measure can identify the people with 90 percent accuracy utmost.
However, the software still has a lot of limitations. First, the team's data set is small-scale, and they have yet to try it on a larger number of people. Second, there is the possibility of misidentification.
This is something that Hassanat and his team want to look into. They will include other information in their algorithm, such as detecting the length and width of a finger.
Additionally, detecting the V hand gesture does not automatically reveal the person's identity, but the tech is still an extremely valuable tool.
The research was submitted to the Cornell University Library.
Photo: Ion-bogdan Dumitrescu | Flickr
