Design and development of a real-time gesture recognition system
Gesture recognition enables a hands-free or device-free interaction with a computer thereby expands the range of possibilities for computer usage. Recognition of the gesture is difficult because of spatial variation and temporal variation among gesture performed by different subjects. A gesture recognition system based on skin detection is developed and presented in this thesis. This system takes the video stream as input, extracts hand area and computes hand motion features, and then recognizes that feature. The process flow of the gesture detection algorithm is as follows. First, it identifies the hand region and calculates the motion trajectories of hand. Then, it creates a motion pattern of the average trajectories for each gesture. Finally, the system classifies pattern based on Euclidean distance and cross-correlation pattern matching techniques. To test and evaluate the algorithm, this system was built using Visual C++. This system is verified for 20 gestures prepared by six performers in different environments. The experiment results show that the average gesture recognition rate is 89% in Cross correlation method and 85.20% in Euclidean distance method. The significance of the system is that it can recognize the gesture feature and display the recognition result in less than 1/30 second. It detects the gestures independent of skin color and physical structure of the performer. The system also works in different complex background with normal intensity of light. ^
Engineering, System Science|Computer Science
"Design and development of a real-time gesture recognition system"
ETD Collection for Tennessee State University.