Study On Dynamic Gesture Recognition On Complex Background

In today’s world, humans and computers is getting closer, human computer interaction has become an indispensable part of daily life. In recent years, A variety of novel human-computer interaction techniques become very active. These studies include face recognition, gesture recognition, posture recognition and tracking, etc.The human gesture recognition is novel human-computer technology based on vision, it got rid of a lot of the shackles of tradition, more in line with the interpersonal habits, to achieve a natural exchange. Among a variety of gestures, the hand gesture is most expressive and the frequently used one, In recent year, many researchers in this area have done a lot of work.In a specific complex background, from identification of all the details, to improve and propose new identification methods, by comprehensive comparison of various dynamic recognition methods and pointing out the shortcomings. The gesture trajectory eigenvectors extracted, first, the skin color segmentation based on YIQ. YUV color space, ordinary frame difference method to segment the moving gesture, the resulting image of the color space to the color image and the frame difference method for intersection, Eliminate skin color face interference. Eliminate the interference of other moving objects in gesture image, regional connection to eliminate noise and fill the gesture area. Based on the above analysis, improve to the gesture trajectory eigenvectors extraction method based on H channel frame difference method, skin color segmentation on YIQ and YUV. and CamShift algorithm. Finally. a detailed analysis of the hidden Markov model theory and signal recognition, combined with gesture recognition and the establishment of a hidden Markov model classifier, about six kinds of simple gestures to train and classify, and achieved a higher recognition rate.