The Command Control of a Two-Degree-of-Freedom Platform by Hand Gesture Moment Invariants

Abstract

In this paper, the functional commands based on hand gesture are designed by the method of moment, which is invariant to the translation, rotation and scale of a hand gesture. After the transform of the original image with a hand gesture into the YCrCb coordinate, the segmentation of the skinlike object is obtained by the thresholds of Cr and Cb. Then the dilation and median filtering and the area constraint are employed to obtain an acceptable binary image. Various feature vectors corresponding to different processed hand gestures are applied to train the input weight matrix and layer weight matrix of a probability neural network for classification. Furthermore different lighting conditions verify the robustness of the image processing and classification. Finally, eight hand gestures are designed as the commands for the motion control of a 2 DOFs platform. The experiment confirms the effectiveness of the proposed method.