The command control by hand gesture with Hu and contour sequence moments and probability neural network

In this paper, the functional commands based on hand gesture are designed by Hu moments and contour sequence moments, which are invariant to the translation, rotation and scale of a hand gesture. First, the original image with a hand gesture is transformed into the color space of YC
r
C
b
. The segmentation of the skin-like objects is obtained by suitable thresholds of C
r
and C
b
. In sequence, the morphological filtering and the shape selection are employed to obtain various acceptable region-based and contour-based binary images of different hand gestures. 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 for Hu moments and contour sequence moments of eight hand gestures are compared to verify the robustness of the image processing and classification.