Good Thinking

Smartphone tech recognizes objects by being knocked against them

A Knocker-equipped smartphone, after being knocked against a smartwatch
KAIST
A Knocker-equipped smartphone, after being knocked against a smartwatch
KAIST

While there are already systems that allow smartphones to identify objects by photographing them, such technology doesn't work well in dim light – plus, not everyone wants to bother taking photos. A new system, however, can identify objects simply by knocking a smartphone against them.

Developed by scientists at the Korea Advanced Institute of Science and Technology (KAIST), the tech is appropriately enough known as Knocker.

It's initially "trained" on sample objects, using the phone's existing microphone, gyroscope and accelerometer to learn the specific sounds and vibrations that occur when the phone makes hard contact with each item. When the phone is subsequently knocked against something, it compares that object's sound and vibrations to those that it's learned, looking for a match. If a match is found, then the object is identified.

Knocker has already been trained to recognize 23 commonly-used items such as laptop computers, bicycles, books, and water bottles. In its current state of development, it has an identification accuracy rate of 83 percent when used in noisy settings such as roadsides or busy cafes, although that number climbs to 98 percent in quiet locations.

Of course, most potential users will already know what the objects that they regularly use are. Instead, it is envisioned that Knocker could be utilized for applications such as ordering new bottled water when the phone is knocked against an empty bottle, noting a user's parking location when the phone is knocked against their locked bike, or turning the lights off when knocked against a bed's headboard.

"This new technology does not require any specialized sensor or hardware," says project leader, Prof. Sung-Ju Lee. "It simply uses the built-in sensors on smartphones and takes advantage of the power of machine learning. It’s a software solution that everyday smartphone users could immediately benefit from."

More possible uses for the technology are demonstrated in the following video.

Source: KAIST

  • Facebook
  • Twitter
  • Flipboard
  • LinkedIn
1 comment
warren52nz
That's pretty cool. It might get annoying if it makes a mistake though. Like predictive text.