Using Configuration States for the Representation and Recognition of Gesture

摘要: A state-based technique for the summarization and recognition of gesture is presented. We define a to be sequence states in measurement or configuration space. For given gesture, these are used capture both repeatability variability evidencedin training set example trajectories. The positioned along prototype shaped such that they narrow directions which ensemble examples tightly constrained, wide great deal observed. develop techniques computing trajectory an trajectories, defining prototype, recognizing gestures from unsegmented, continuous stream sensor data. approach illustrated by application range gesture-related sensory data: two-dimensional movements mouse input device, movement hand measured magnetic spatial position orientation sensor, and, lastly, changing eigenvector projection coefficients computed image sequence.