Advanced Spatial Recognition
Professor Lee Sung-hee at the Korea Advanced Institute of Science and Technology (KAIST) announced on October 20 that his research team developed an AR avatar motion generation technique for simultaneous user motion imitation and spatial recognition.
An AR avatar can be defined as an one present in a virtual or remote space as a representative of its user and imitating his or her motion. Existing AR avatars are characterized by being unable to recognize the difference between the user’s space and the avatar’s space and simply passing through objects present in the virtual or remote space, which makes them not so realistic.
In order to overcome this limitation, the research team analyzed environmental information by making use of a depth-based camera, modified its avatar’s motion based on the information, and enabled it to adapt itself to altered spaces. In an experiment, the research team placed a kitchen chair in the user’s space, placed a single couch in the remote space, and then succeeded in making the avatar sit in the couch by performing a motion and taking a posture suitable for the couch. This means the avatar away from its user is capable of reproducing its user’s motion without any distortion in a given space.
“This technique can be utilized in the field of social telepresence, which enables people away from one another to communicate as if they were in the same space,” the professor explained. Details of the research were presented at this year’s International Symposium on Mixed and Augmented Reality (ISMAR) in Merida, Mexico.