Research On Gesture Recongnition Based On Multi Feature Fusion

With the development of computer technology,the way of human-machine interaction has been constantly innovating.From traditional mouse and keyboard to the current popular touch screen,to more advanced voice control,human-machine interaction technology is more and more humane.Comparing with traditional human-machine interaction,gesture interaction has the advantages of non-contact operation,simple learning and portability.Therefore,gesture based human-machine interaction technology is of great significance.Gesture recognition is one of the basic research issues of gesture interaction.It is the focus of the majority of researchers.It is also a difficult problem to be solved in the research of behavior recognition.The main work and contribution of the paper are as follows:(1)Hand gesture segmentation based on complex background environment.The gesture contains a large number of interactive information consistent with human cognitive habits.In view of the complex background and external factors(such as illumination,occlusion and motion)in gesture segmentation,we proposed a gesture segmentation method by skin detection based on YCbCr color space.The methods of skin detection based on elliptical model is used first to segment the skin regions under complex background environment.Then the Otsu method is employed to obtaining segmentation threshold,turn out we obtain a high quality gesture binary image.The experimental results show that the proposed algorithm is robust to complex background environment,and can achieve more efficient gesture segmentation.(2)Gesture recognition based on multi feature fusion.Compared with traditional methods,image recognition based on convolutional neural network has more accurate recognition accuracy and speed.Therefore,a gesture recognition method based on multi feature fusion is proposed in this paper.First,the gesture features of skeleton and edge are extracted from the segmented hand gesture images,and then the convolution neural network is utilized to recognize the extracted gesture.The experimental results show that the algorithm is robust against complex background environment,and the proposed method has better recognition results under different data sets.