Summary: | 碩士 === 國立宜蘭大學 === 資訊工程學系碩士班 === 104 === Body-behavior is one of the most important communication in daily life. however, the rapid development of technology, human and computer interaction is born out. Nowadays, the prototype of Human-Computer Interactive Mode is universal development, but many still need to be wearable equipment to detect and track. Therefore, the objective of this research is to develop a system that Posture Interactive Learning System Based on Human Action Tracking. Here, to track human physical action, it is determined whether the video image is in the same action as the benchmark score. Our system includes: Image Preprocessing, Feature Extraction, the operating position of hands to judge, Trajectory Tracking, Chain Code and scoring mechanism.
The core techniques involve a rule-based approach for partial posture action, the Segmentation of Hand and Body, Skeleton Tracking, Body Feature Extraction and Alignment Similarity. The experimental results demonstrated that our system is able to successfully segment the gesture and physical action. And then each action of the similarity comparison can be identified and scored, its left and right hands when cross-interactive in different posture is be distinguished. and to re-trace the upon trajectory changes, further the action can repeatedly practice and remind to mistake.
In conclusion, the main purpose of this research is for the teaching mode of immediate virtual coach. Our system could be incorporated ultimately in the computer user-interface that uses a video camera as input device. In addition, its may offer a cost-effective solution to other motion capturing systems that often require markers or special motion sensors.
|