View invariant gesture recognition using 3D motion primitives.
Michael B. HolteThomas B. MoeslundPublished in: ICASSP (2008)
Keyphrases
- view invariant
- motion primitives
- human motion
- gesture recognition
- human actions
- video sequences
- image sequences
- action recognition
- human motion analysis
- multi view
- single view
- spatio temporal
- three dimensional
- motion estimation
- motion capture
- hidden markov models
- human body
- motion parameters
- motion capture data
- human computer interaction
- hand gestures
- optical flow
- space time
- viewpoint
- computer vision
- motion trajectories
- body parts
- low level
- activity recognition
- structure from motion
- markerless
- moving objects
- high level
- feature points
- humanoid robot
- multiple views
- visual features
- multiple viewpoints