Hand region extraction and gesture recognition from video stream with complex background through entropy analysis

Hand gesture recognition utilizing image processing relies upon recognition through markers or hand extraction by colors, and therefore is heavily restricted by the colors of clothes or skin. We propose a method to recognize band gestures extracted from images with a complex background for a more natural interface in HCI (human computer interaction). The proposed method obtains the image by subtracting one image from another sequential image, measures the entropy, separates hand region from images, tracks the hand region and recognizes hand gestures. Through entropy measurement, we have color information that has near distribution in complexion for regions that have big values and extracted hand region from input images. We could draw the hand region adaptively in variable lighting or individual differences because entropy offers color information as well as motion information at the same time. The detected contour using chain code for the hand region is extracted, and present centroidal profile method that is improved little more and recognized gesture of hand. In the experimental results for 6 kinds of hand gesture, it shows the recognition rate with more than 95% for person and 90-100% for each gesture at 5 frames/sec.