METHOD FOR EXTRACTING HAND FEATURE BASED ON CURVATURE ANALYSIS FOR RECOGNITION OF VARIOUS HAND GESTURES

摘要

The present invention relates to a method for extracting a hand feature based on a curvature analysis for recognition of various hand gestures. The method for extracting a hand feature based on a curvature analysis for recognition of various hand gestures comprises: (a) a preprocessing step of performing a hand candidate group extracting process of extracting a candidate group presumed as a hand region so as to find the hand region from a lot of objects in an input image and a process of finding the hand region from the hand candidate group; and (b) a step of extracting the hand feature from an image of the hand region, that is, a feature extracting step of extracting a hand outline, a boundary point between fingers, an outline feature point, and a central point of a hand and extracting feature information on the number of opened fingers and a state of fingers stuck attached. A hand feature extracting algorithm based on a curvature analysis capable of recognizing the number of fingers and the attaching of fingers is used for extracting a feature necessary for the hand gesture recognition. The hand feature extracting algorithm recognizes various hand gestures by detecting the hand region from the input image through a skin color range filter and a labeling based on a color model and extracting the feature about the number of the opened fingers and the attached fingers by using the outline, the feature points, and curvature information extracted from the outline and the feature points. According to a test result, while a recognition rate and a processable frame rate are similar to those of the existing algorithm, the number of cases of gestures which can be defined with the extracted feature is about four times more than the number of cases of gestures in the existing algorithm. Therefore, much more various hand gestures can be recognized.;COPYRIGHT KIPO 2016