This paper proposes a novel method for detecting hand-raising gestures in meeting room and classroom environments. The proposed method first detects faces in each frame of the video sequence in order to define the region of interest (ROI). Then the system locates arms in the region of interest by analyzing the geometric structure of edges on the arm instead of directly detecting the hand. The location and the orientation of a detected arm respect to the location of the face is used to make a decision on whether or not a person is raising hand. Finally, the frequency of a raised hand detected in previous frames is used to eliminate false positive detections and robustly detects persons who are raising a hand. Unlike major visual gesture recognition systems, our method does not rely on skin color or complex tracking algorithms, while achieving 92% sensitivity and 92% selectivity.
展开▼
