One of the challenges of metaverse is to identify a person after the aesthetic changes that can occur there, for this reason, a group of researchers developed a system to recognize another, without the need to scan their face, all that is needed are the body movements.
Those in charge of this project are scientists from the University of Berkeley, the RWTH Aachen in Germany and the company Unanimous AI.
“We have known for a long time that movement reveals information about people, but what this study recently shows is that movement patterns are so unique to an individual that they could serve as biometric identification, just like face or hand recognition. fingerprints,” Vivek Nair, one of the authors of the research, told The Register.
Biomechanical recognition is just as effective as facial recognition. (Unsplash)
How identification works
According to the researchers, studies dating back to the 1970s show that people can recognize a friend based on their body movements. So they took that idea as the basis of the project to determine how feasible it is to take it to the metaverse and know who is behind the avatar.
biomechanics It is a characteristic that each person has, just like footprints or their face, which is why it was one of the techniques used in the project.
To achieve these results, they took 55,541 users, with whom they trained the recognition model for 5 minutes per person, collecting all possible data. However, they assure that for a company with a large budget this process would be much more efficient.
After the training they were able to identify the analysis in 10 seconds with 73.20% accuracy, but when they extended the time to 100 seconds it rose to 94.33%. So, in just over a minute and a half, they were able to recognize a person in the metaverse just by seeing the movements of their hands and head, without seeing the face.
The virtual space they used for the experiment was the Beat Saber game and the Meta Quest 2 glasses. In addition, they obtained a total of 2.66 million recordings of users playing, to make up a dataset of 3.96 TB after 713,000 game sessions.
Biomechanical recognition is just as effective as facial recognition. (FreePik)
It may interest you: More than half of the people in the world want to watch movies and play in the metaverse
This large amount of data was key because they needed to look at various aspects of the body such as the orientation of the head and hands, the timing in which they reacted to in-game stimuli, performance, and accuracy.
Information that can be obtained thanks to the advancement of virtual reality glasses that allow evaluating the height, height, length of the arms and the distance between the eyes that provides particular features of a person.
It may interest you: They create the first metaverse that rescues Afro-Colombian culture and traditions
“This really changes the way we think about the notion of ‘privacy’ in the metaverse, because just by moving around in VR…you could be identifying yourself!” Nair said.
Finally, the project opens the doors to new forms of recognition within the metaverse, although it also evaluates the privacy policies of the users in these spaces.
