Researching On Gesture Recognition And Its Application

The gesture recognition technology based on computer vision is an important research direction in the field of human-computer interaction.With the rapid development of 3D somatosensory technology,under the natural interaction interface,the main mode that gesture recognition technology applied in HCI system is a short range control of the operating equipment.The research of gesture recognition and tracking method based on Kinect,mainly includes four aspects such as gesture tracking,gesture image segmentation,feature extraction and gesture recognition algorithm.The mainly researches are as follows:(1)Researching on gesture tracking method based on depth information,Kinect2.0 is used as an input device for depth data.The spatial position of human bone joints can be tracked with skeleton tracking technology.Due to serious noise interference in the depth of the original histogram,data fusing the depth value and the neighborhood depth value is put forward for statistics.Then a 2D depth histogram based on depth value and neighborhood depth value is proposed for reduce noise interference.According to the statistical results of histogram,the Otsu algorithm is used to obtain the target area of the gesture image.(2)A static gesture recognition algorithm of DAG-SVMs with the moment of the Hu moment and the edge length is proposed in this thesis.Firstly,the edge contour of the gesture image is detected by Sobel operator.Secondly,as the new features of the static gesture,the Hu moments are combined with the edge length moment for recognition.Multi-class support vector machines(DAG-SVMs)are utilized for machine learning,and the recognition accuracy is improved by optimizing the random structure of the algorithm.The simulation results show that the recognition rate of the algorithm is up to 97.69%.Compared with the algorithm which only use the single feature Hu moments,the recognition rate of the algorithm proposed in this thesis increased by 1.78%.Based on the results of static gesture recognition,a multimedia volume control system by static gestures is designed in this thesis.In the system the results of static gesture recognition can be converted to different instruction in order to control the multimedia volume.(3)Dynamic time Warping(DTW)algorithm is utilized for dynamic gesture recognition.Under the continuous time sequence,as dynamic hand gestures features,geometric parameters and quaternion features are extracted for constructing feature template.In order to identify six kinds of dynamic gestures,the DTW algorithm is utilized to find the optimal matching path between test samples to the reference templates.Simulation experiment results demonstrate that the recognition algorithm presented in this thesis can effectively identify six kinds of common dynamic gestures and the recognition rate is 94.73%...