 |
Email:
|
|
| Beschrijving: |
The field of pose, action and gesture recognition is relatively young and changes very rapidly. Commercial applications have seen much progress recently with for example the Kinect motion sensor. In more academic research progress has been made in several directions. One is accurately detecting someone’s pose and movement from markerless video feeds. Another is for example using someone’s captured movement to recognize, to some degree of precision, their gender or identity.
Not many studies, however, concern themselves with the meaning of the gathered pose data, reducing gesture recognition to a very sophisticated way to push a button (“left”, “right”) or limiting the detection to surveillance purposes. The University of Twente and re-lion – an Enschede based company that builds fullbody motion capture virtual reality simulations for, mainly, military infantry training – have committed themselves to the Gate Knowledge Transfer Project “Computer Animation for Social Signals and Interactive Behaviors”, which tries to take this concept one step further and ask: what is the social signal in the detected pose or motion? This social signal is then used as input for the simulation to, for example, steer artificial intelligences.
In my Master's research I will take a first swing at this compex issue. I will try to detect non-verbal signals using the motion capture suit and software system of re-lion so that further research within the Gate project can build upon mine. The particular use-case that I will work on is the detection of arm gestures that the military uses to silently issue commands. Being able to detect such gestures in real-time in the re-lion virtual reality simulation would allow for novel ways of interacting with the system, for example enabling the user to communicate non-verbally with artificially intelligent agents in the system, using gestures to steer the scenario in some direction or mark a point in time with a gesture for the After Action Review.
The technological challenge lies in detecting wildly varying gestures in real-time in a very realistic use-case: users of this system do not sit still while making gestures, instead they move around a simulated VR environment, crouch, run, shoot at enemies and evade virtual objects. All this takes accurately detecting arm gestures to a new level of complexity.
|
|
| Beschrijving: |
This paper tries to defi ne Augmented Reality from a Human Media Interaction perspective and gives an exploration of diff erent uses of Augmented Reality in contemporary applications. Finally it presents a simple Augmented Reality display that I developed for virtual meetings.
I conclude that my display does not yet fully ful fill the requirements that I defi ned for being considered Augmented Reality, being the need for some interaction with the real world and a slight perceptual mismatch that triggers the user to view it as something "magical". |
|
|
|
