快猫短视频

Your phone may be able to recognise you just from the vibrations in your hand. An algorithm trained to identify people this way did so with more than 90 per cent accuracy.

at the Toulouse Institute of Computer Science Research in France and his colleagues trained an artificially intelligent聽algorithm to identify an individual based on patterns in their hand vibrations when holding a聽smartphone.

The team recruited 217 volunteers to each hold a smartphone while the device鈥檚 sensors collected data during either a single 30-second measurement session or a series of such sessions. The AI used about half of this data to figure out the best way to differentiate individual smartphone users.

The team then tested the algorithm with the other half of the data to see if it could identify specific individuals in the group. Across 1100 tests, the algorithm identified the correct person around 92.5 per cent of the time within 1.5 seconds.

The system could eventually be used as a form of password. One of the main problems for it at the moment is that other movement affects the accuracy. 鈥淚f you are in a car or train, for example, [鈥�] the vehicle motion will essentially drown out the signal,鈥� says at the University of Lille in France, part of the team.

Other researchers have faced similar challenges when studying signals like those used for monitoring聽heart health, says聽 at the Georgia Institute of Technology. He also says that if the system was to be used as a way to authenticate who was holding a device, the accuracy would need to hold up while users are undergoing heavy or stress that could subsequently affect the vibrations in their hands.