The Research Of Gesture Recognition Based On Depth Image From Kinect

The next ten years, with the continuous advances in computer science and technique,human-computer interaction model will continue to change. Compared to using the keyboard,mouse and other media equipment, gestures in human-computer interaction more natural andconform to human behavior habits. So, the gesture recognition technology has become the hottopic in the field of human-computer interaction research, and has wide prospects and hugemarket potential.The traditional vision-based gesture recognition using ordinary webcam as image outputdevice, but it is extremely vulnerable to changes in light and skin color interference, and therecognition results are often not up to people’s expectations hope. In this paper, we use thelatest high-tech products that Kinect depth camera, it can avoid the impact of environmentalfactors on gesture recognition. In the static gesture recognition, this paper proposes a fingertipdetection method that is depth-based convex defects detection, it uses depth information andneighborhood features to segment the hand area, and selects the canny operator to extractpalm’s contour. In the fingertip detection process, it can detect all the possible fingertip areasthrough the convex hull, and then rely on the relationship between the convex hull and defectsfound real fingertips. In the dynamic gesture recognition, it estimates skeleton joints’ actualposition from depth image obtains by Kinect sensor, and after normalization treatment, itextracts six joints’ trajectories as the feature vectors. This paper analyzes the dynamic timewarping algorithm (DTW), by limiting the search range to complete the globaloptimization of DTW algorithm. It can shorten the matching time and realize the recognitionof six kinds of dynamic gesture tracks, the average recognition rate of96.3%. In order toverify the feasibility of the algorithm, it is applied in the PPT operating system, to achieve theuser in the non-contact conditions on PPT playing operation and control page. Theexperimental results show that in the light of changing and complex environment, havereceived good recognition results, so that the PPT operating system has strong robustness.