Motion-Pose Recurrent Neural Network with Instantaneous Kinematic Descriptor for Skeleton Based Gesture Detection and Recognition.
Zhi ZhangYonghong SongYuanlin ZhangPublished in: ACPR (2017)
Keyphrases
- recurrent neural networks
- hand gestures
- human hand
- degrees of freedom
- partial occlusion
- gesture recognition
- action recognition
- object recognition
- joint angles
- hand postures
- motion history images
- neural network
- recognizing actions
- inverse kinematics
- object detection
- end effector
- pose estimation
- sign language
- feed forward
- human motion
- position and orientation
- complex valued
- fine grained categorization
- recurrent networks
- hand motion
- hand pose
- reservoir computing
- echo state networks
- human movement
- hidden layer
- recognition algorithm
- human computer interaction
- parallel robot
- d objects
- orientation histogram
- activity recognition
- motion planning
- human detection
- configuration space
- neural model
- feature extraction
- kinematic constraints
- artificial neural networks
- camera motion
- human pose
- expert systems
- motion patterns
- computer vision
- image sequences
- loop closure
- optical flow
- hidden markov models
- complex background
- robot motion
- articulated objects
- kinematic model
- parallel manipulator
- body posture
- motion capture
- human body
- mechanical devices