Automatic Hand Gesture Segmentation Based on Multi-Feature Criteria

摘要

Detecting meaningful hand gestures in real-time with multi-variant room conditions presents many challenges. Variations such as room lighting, the detection for the presence of skin color, and determining the meaning of the hand gesture, if one exists, must be resolved for an automated hand gesture detection system. Solving these problems would contribute toward a non-verbal communication system that could benefit a variety of people and organizations who utilize hand gestures for communication. For the initial approach for this automated system, a multi-feature, three criteria human computer interface consisting of an object moving into a predefined three dimensional space, the presence of skin color, and non-motion is presented. A systematic method was developed to distinguish paused motion from hand movements from RGB and depth images so that pattern recognition techniques can be effectively utilized to interpret the hand gesture. The approach was successfully validated by experiments.