Research On Non-contact Hand Motion Recognition

In the recent years, the technology of the computer vision has beendeveloped rapidly. Human-computer interaction, as a branch of the computervision, is a very important research field. Except the language, Hand gesture isthe second communication way people normally choose. Especially there isdifficult to communicate by language. Non-contact hand motion recognitionmeans that people only recognize by video equipment rather than other detectionequipment. The paper is to take the single hand motion under the normal videocamera as the research object, with the main content as following:Firstly, it introduces the theory and algorithm related to the motionrecognition. Regarding the noise from the motion object detecting, it makesimprovement based on the commonly used image denoising algorithm. And thenit process with the method by threshold area elimination&hole filling algorithmon the basis of edge profile.Secondly, it takes the skin color as the feature to detect the location of thehuman hands and to collect the track of the hands motion. In this paper, thehands’ static state is taken as the sign of the end of the moving sequence. Aspeople have different habit, the hands speed may be near to static when themoving direction is changed. However, the points which move slowly can not beconsidered static state simply. For this situation, the paper uses extended finitestate machine theory to determine the hands static state.Finally, for the situation with the same hands moving trajectory but withdifferent scale, it recognizes the hands moving trajectory by using Hausdorffdistance which has the transformation characteristics of translation and scale.