Automatic action segmentation and continuous recognition for basic indoor actions based on kinect pose streams.
Yun HanSheng-Luen ChungShun-Feng SuPublished in: SMC (2017)
Keyphrases
- recognizing actions
- action recognition
- human actions
- activity recognition
- continuous action
- action detection
- fully automatic
- action space
- pose estimation
- symbol recognition
- human activities
- human pose tracking
- features extraction
- segmentation algorithm
- robust recognition
- segmentation method
- body motions
- action language
- depth cameras
- recognition rate
- recognizing human actions
- numeral strings
- action selection
- computer vision
- image segmentation
- action rules
- action sequences
- human pose
- recognition algorithm
- reasoning about actions
- level set
- joint action
- medical images
- improve recognition accuracy
- concurrent actions
- partial occlusion
- human body
- intensity images
- d objects
- state space
- mobile robot
- object recognition
- data streams
- pose invariant face recognition
- depth images
- video sequences