CMU Develops An Algorithm To Help Robots Understand Social Groups
@pensu-8tNeGU
•
Oct 23, 2024
Oct 23, 2024
1.4K
How do you know what people are interested in? A simple answer would be, depends on what are they looking at. And how do you except a robot to learn that? The research team at Carnegie Milan University (CMU) has the answer to that question. The researchers has developed an algorithm which use head-mount cameras to track what people are gazing at and where there sighting intersects.
 The idea behind the research is to find out different social groups, which is a tedious task otherwise. The method was tested by providing some people with head mounted cameras, based on where their gazes converged in three dimensional space, the team was able to determine were they listening to same speaker or talking in a group or watching a game together?
#-Link-Snipped-#
The “algorithm for determining social saliency†can be used in various application including knowing people interest, reading their body movements etc. According to Hyun Soo Park, a Ph.D. student, who developed the algorithm with Yaser Sheikh and Eakata Jain, “This really is just a first step toward analyzing the social signals of peopleâ€. The only issue in this whole scenario is the head-mounted cameras, but team hopes that the head-mounted systems, like the ones with integrated eye glass frames are going to be more common in future. More information about the project is available on <a href="https://www.cs.cmu.edu/~hyunsoop/gaze_concurrence.html" target="_blank" rel="nofollow noopener noreferrer">Gaze Concurrences</a>.
 The idea behind the research is to find out different social groups, which is a tedious task otherwise. The method was tested by providing some people with head mounted cameras, based on where their gazes converged in three dimensional space, the team was able to determine were they listening to same speaker or talking in a group or watching a game together?
#-Link-Snipped-#
The “algorithm for determining social saliency†can be used in various application including knowing people interest, reading their body movements etc. According to Hyun Soo Park, a Ph.D. student, who developed the algorithm with Yaser Sheikh and Eakata Jain, “This really is just a first step toward analyzing the social signals of peopleâ€. The only issue in this whole scenario is the head-mounted cameras, but team hopes that the head-mounted systems, like the ones with integrated eye glass frames are going to be more common in future. More information about the project is available on <a href="https://www.cs.cmu.edu/~hyunsoop/gaze_concurrence.html" target="_blank" rel="nofollow noopener noreferrer">Gaze Concurrences</a>.