Gesture Recognition Based On Kinect And Its Application In Manipulator Control

With the rapid development of robotics,robots have been widely used in various fields such as life and production,which makes the interaction between humans and robots a hot research topic.However,the traditional robot control methods are monotonous and cumbersome.The interaction between people and robots is not intuitive and inefficient,which puts higher requirements on new human-computer interaction.The gesture-based interactive system not only has academic research value,but also has a broad application prospect.Based on the RGB-D image information collected by the Kinect depth sensor,the paper implements the static gesture and dynamic gesture recognition respectively,and applies the gesture recognition technology to the control of the manipulator.It realizes the gesture control of the uArm manipulator to write digits and complete the verification of gesture recognition to the intelligent control of the manipulator.In the study of static gesture recognition,the hand gesture segmentation method combined adaptive depth threshold with skin color detection is adopted in this paper,which can overcome the interference of light and segment gesture area from a complex background effectively.After morphological processing and smoothing on the gesture region,the Border-Following algorithm is used to extract the gesture outline.Finally,using the calculated multiple feature parameters based on gesture outline as classification nodes,a static gesture recognition classification decision tree model is established,which enables fast and accurate identification of custom static gestures.Because of the adaptability of feature parameters and the invariability of scaling and rotation,the static gesture recognition method has good robustness.For dynamic gesture recognition,a dynamic gesture start-stop detection method based on a dynamic sliding window is proposed to extract valid gesture sequences accurately.We encode the effective gesture sequence by improved Freeman chain code,and use the improved dynamic time warping(DTW)algorithm to match the effective gesture sequence with gesture template library to realize dynamic gesture recognition.The improved DTW algorithm can effectively prevent the recognition errors caused by deformation of the template by controlling the slope to limit the search range.Moreover,the method of table lookup instead of repetitive boundary calculation reduces the time-consuming of the algorithm,which improves the correct recognition rate and recognition efficiency of dynamic gestures.After implementing static gesture recognition and dynamic gesture recognition,the gesture recognition results are converted into executable commands to realize intelligent control of the manipulator.