TV remotes could soon be in tune to viewers’ interests with the help of two ASU students.
Mariano Phielipp, a 2008 ASU alumnus who received a doctorate in computer science, helped develop a way remote controls can identify people based on their TV preferences and how they hold the remote. This could help advance research to customize the TV viewer experience.
The algorithm tracks how a remote is held and which buttons are pressed and begins working to identify the user, said Phielipp, who now works for Intel’s Digital Home Group. It should then be able to pick up on the channels they like to watch.
Magdiel Galan, a computer science doctoral student, interned with Intel for four months in order to help with the project, though he is not currently working for them.
Phielipp said Galan helped research and test different approaches to make the algorithm work the way they wanted. Galan was involved in all of the collaboration and helped with writing the final paper.
The results were presented at the Artificial Intelligence for Interactive Digital Entertainment conference in Atlanta in July.
The goal of the research is to create a more customized experience for the user, Phielipp said. Researchers hope the remote can customize preferred television channels, audio channels or online games played through the television for each user.
This research can be compatible with different televisions as long as they can run the specific algorithm, Phielipp said.
The research done by Intel is only for research purposes unless a buyer wants to purchase the chips they have developed and tested in remotes, he said.
There are no plans currently for a product to be made, said Jeffrey Hightower, a researcher at Intel, but in the future it should make watching television better and more personal.
In 2009, once the algorithms were determined to have some accuracy, the project moved on to the next step, which included Phielipp’s work.
Phielipp’s research was a continuation of research by Hightower and other members of the Intel Lab. Lab members figured out a way to collect the data and make it capable of classifying the user.
Hightower said it took about six months to do his initial research before it was passed along to Phielipp and the product team.
To test the algorithms and make sure they were doing what Phielipp and his other researchers wanted, the data was collected over three weeks from four different families.
The families were different in order to give a variety of data. This testing showed how accurate the algorithms were in improving the user’s experience, Phielipp said.
Reach the reporter at hfogel@asu.edu
