Gesture Recognition Based On Kinect And Human-computer Interaction

In recent years,with the rapid development of computer technology,human-computer interaction methods have also developed more and more diverse.Compared with traditional human-computer interaction methods using keyboards,mice,and other hardware,the method of using gestures to conduct non-contact human-computer interaction is more direct and natural,and more in line with people's habits in life.Therefore,researchers began to study gesture-based non-contact human-computer interaction methods.By using the image acquisition device to collect user's gesture information,the acquired gestures are identified and the recognition results are used for human-computer interaction.Kinect can not only provide color data,but also can acquire depth data corresponding to color images,providing convenience for visual-based human-computer interaction.The paper selected several commonly used static gestures and dynamic gestures as basic recognition targets.In the gesture preprocessing stage,according to people's habits of gesturing and depth information provided by Kinect,quickly locate the depth range of the hand,and use the improved block search method to select a block of a suitable size to quickly find a point on the outline of the gesture.Using the direction search method to obtain gestures and their outline information in the depth image.In the static gesture recognition,the positions of the palm and fingertips are obtained,and the relative positions of the fingertip points and the palm and fingertip points are extracted as features of the static gesture,and a static gesture is described with 10 frames of continuous images.When using the dynamic time warping algorithm to identify,add diagonal weights to ensure that the diagonal movement,resulting in accurate recognition results.In dynamic gesture recognition,the motion trajectory of the fingertip point is represented by a vector and then encoded to describe the dynamic gesture.In the gesture recognition,the dynamic time warping algorithm is improved,when the algorithm finds the optimal bending path Add the step element,calculate the cost of the curved path under the out-of-sync number,consider the different situations comprehensively,compare the gesture to be recognized with the template gesture,and select the gesture template with the highest degree of similarity as the recognition result of the dynamic gesture.The paper only uses the depth information provided by Kinect to reduce the effects of lighting and complex backgrounds.The improved block search method can quickly and accurately obtain gesture information,thereby accurately extracting gesture features.A reasonable feature is selected to describe the static and dynamic gestures,and the static gestures 1-6 and the dynamic gestures 0-9 are identified,and a higher recognition rate is obtained.Finally,this method is applied to the human-computer interaction,that is,the gesture recognition result is used as a control command to control the 3D printing robot InMoov.In terms of the result,this method has a good application prospect.