The Research Of Gesture Control Andrecognition Algorithm

Gesture control and recognition are the important parts of intelligent human-computer interaction, which are the hot spot in the present study. There are two aspects about the gesture control and recognition. The one is to establish the virtual hand model and the control mechanism. The other is the gesture detection and recognition methods. The research of the gesture control and recognition will contribute to implement the intelligent human-computer interaction.We have conducted the thorough research about the existing algorithms of gesture control and recognition, and do the following work for further study.(1) The hand muscle model which meets the physiological characteristics of the human is established, and the control mechanism based on multi-curve spectrum is proposed. First, we established the hand muscle model based on the hand anatomical structure and the geometric proportion of the joint, and the relationships between finger movement angle and the amounts of muscle contraction are given. Then on this basis, we put forward the control mechanism based on multi-curve spectrum. The multi-curve spectrum can describe movements in two angles of time and space. And it can be express with simple functions. So it can be used to control the movements of virtual hand. Finally, combined the feature of hand movements with the concept of movement constraint of the multi-curve spectrum, we simulate hand movements. The experimental results show that the control mechanism not only reduced the control complexity of the model, but also made the simulation of dynamic gesture further improve.(2) A new detection method of gesture is carried out based on different color space, which can detect gestures accurately and rapidly. The existing methods of gesture detection have some drawback in the highlights and lowlights. Therefore, by analyzing the distribution of a variety of skin color in the color space, we use the quadratic discriminate method to distinguish illumination environment of the input image. According to the brightness of the image information, we divide the image into normal and abnormal light. Under the normal light, the intersection of detection results of YCb Cr and YCg Cr skin color is the result of the gesture detection. Under the abnormal light, we use HSV and YCg Cr color model to confirm the gesture area. The experimental results show that the algorithm can effectively avoid the influence of the brightness information.(3) We put forward a static gesture recognition algorithm based on the principal axis of inertia and a dynamic gesture recognition algorithm based on the multi-curve spectrum. In view of the deficiency in the invariant about rotation and scaling, a static gesture recognition algorithm based on the principal axis of inertia is provided. Using the image centroid and principal moments of inertia, the 8-dimensional joint gesture characteristic is obtained. The experiment results show that the real-time is good, and the average recognition accuracy can reach more than 99%. In view of dynamic gestures, the recognition algorithm based on the multi-curve spectrum is proposed. By tracking the key position of gestures, we get the corresponding multi-curve spectrum. Because the multi-curve spectrum can describe the process of gestures in two angles of time and space, our algorithm not only ensured the real-time of the recognition, but also predicted the gesture.