Research And Application Of Gesture Recognition Based On Depth Image

With the continuous development of science and technology, human-computer interaction is moving towards a simpler and more humanistic direction. Gesture recognition as a very important way of human-computer,In this paper,a gesture recognition method based on depth information is studied by using Kinect sensor.We study the static hand gesture recognition based on depth image, The hand contour extraction using moore-neighbor algorithm and hand gesture recognition using fingertip detection method. This paper presents a curvature detection algorithm based on convex hull, The algorithm based on hand shape characteristics, scanning and filtering the contour points, then making convex hull detection, The advantages of the convex hull detection is larger curvature point defects will not be detected,overcome the error that contour edge is not standard, greatly simplify the clustering operation and the desired fingertip is obtained by curvature threshold.Aiming at the traditional dynamic gesture recognition, taking into account the difference in user size and location. This paper standardized the extraction of skeleton nodes. The normalized position is realized by transforming the Descartes coordinate system into a spherical coordinate system, and the normalized user realizes the distance between the head node and the torso node as the normalization factor. and proposed a DTW restricted algorithm based on weighted multi dimension.In the process of sequence matching, The paper restrict the search path in diagonal direction of the distance matrix and redefine the Euclidean distance between points.In the calculation of the distance between the features, According to the different contribution of each joint point to the dynamic gesture, it gives different weights to improve the recognition accuracy.The above algorithms are proved by experiments and analysis, the algorithm has improved greatly in recognition rate and real-time fields, We apply the algorithm to to Jiazhuang design system, What the keyboard and mouse can do is achieved by gesture.