At this technological age, WiFi is everywhere. And being practically an abundant resource, researchers from the University of California, Santa Barbara, aimed to look for a way to use it other than for internet connection.
In a recent publication in ScienceDaily, the researchers present how their technology could be used to identify criminals entering a building based on WiFi signals and a pre-existing video footage. Their technology, called XModal-ID but pronounced cross modal ID, works by gauging how one's gait affects the signals from two WiFi transceivers. The pattern created from the interference of these signals would then be compared with the extracted signal from a video footage of a certain suspect.
Lead researcher and professor of electrical and computer engineering at the University of California, Santa Barbara, Yasamin Mostofi, explained the fundamental basis of their approach. She said that the proposed approach makes it possible for users to determine if a person behind a wall, that is to say inside a building, is the same as the one in a pre-existing video footage. And this can be done using only a pair of store-bought WiFi transceivers. The technology would only need to utilize received power measurements of a WiFi link. So it does not necessarily require any data of the person to be identified, or any knowledge of the operation of the intricate internals of the system.
In a demonstrative video below, the researchers explain how they converted a video input to an animated 3D mesh that illustrates the relationship of the shape of an individual's body with time, while walking. Using electromagnetic wave approximation, the 3D mesh would then be used to create a WiFi signal simulation that would serve as basis for identifying a person based on their gait.
During application, time–frequency processing would be used to extract gait features from the previously simulated signal and from a real-time WiFi signal. These two would allow the user of the device to compare them and determine if they match.
Testing their device, the researchers found that it was able to identify walking persons correctly at a rate of 84 percent. In a statement, Mostofi explained that identification is not exactly a simple task. "However, identifying a person through walls, from candidate video footage, is a considerably challenging problem," she said. She goes on to say that each person's move is unique.
The researchers will be presenting their technology and findings in the 25th International Conference on Mobile Computing and Networking (MobiCom) to be held in Mexico this October.
