The development of “smart TVs” is also leading, not surprisingly, to the creation of a smart remote as well, thanks to Intel .
At the annual Association for the Advancement of Artificial Intelligence, Intel researchers presented a paper detailing research into a television remote that can sense just who in a family is using the remote based on how they physically handled it and pressed its buttons.
Researchers took accelerometer technology, which is used in consumer devices such as iPhone to shift the screen to a viewer’s position, and attached it to a TV remote. The accelerator provides information about motion. For instance, was the remote picked up by a left or right hand? Data from button presses usage, such as press duration, is also collected.
This information is analyzed by algorithms that look for patterns to help identify users. Testing was conducted in real-life, on households that agreed to have their usage recorded by camera so the results could be checked.
It was tested on five households of two to four people over a several week period, producing a variety of ranges in accuracy — 70%-80% and above, all depending on the type of test, according to the paper. Work continues on refining it.
Mariano Phielipp, a researcher in Digital Home Group at Intel, who presented the paper, said the potential uses could be automatic sign-on to a game, for instance, or prioritizing of channels, depending on who has the remote.
Issues remain for establishing identity, said Phielipp. “It doesn’t mean that we have solved all the identity problems, but it is one way of helping the TV recognize who you are,” he said.
Although Intel isn’t in the business of making consumer devices, it is doing “a lot of work in the TV area, Smart TV especially, and working with a variety of OEMs,” Intel spokesman Bill Calder said by email.
Of the smart remote, highlighted in a paper titled, Fast, Accurate, and Practical Identity Inference Using TV Remote Controls (PDF document) , Calder said that “for now this is just research, to see if it can be done.”
The paper’s authors include Phielipp abd Magdiel Galan of the Dept. of Computer Science and Engineering at Arizona State University, and from Intel: Richard Lee, Branislav Kveton, and Jeffrey Hightower.