Research On Non-contact Interaction Technology Based On Depth Information

In the current development situation of human-computer interaction,interactive information input methods are evolving from the traditional mouse,keyboard,to touch screen,voice,gestures and other simpler interactive methods,which are gradually changing to natural interaction.Among them,human gestures are flexible,intuitive,and capable of representing rich and varied information,and it is an important method to achieve non-contact human-computer interaction.The Kinect sensor can obtain depth information of the recognition scene,and implement non-contact gesture interaction conveniently.This paper proposes an improved threshold segmentation method for the demand of accurately segment gestures at different distances in real-time.First,it uses bone tracking to coarsely locate gestures,and then uses spatial stratification detection to locate the depth position of the gestures combining with skin color information.At last,position gestures accurately based on gesture contours.Experimental results show that the threshold segmentation method proposed in this paper can accurately segment gestures at different distances in real-time,and unaffected by location,complex background.In gesture recognition,the k-cosine curvature method of the graphical algorithm and the CNN classification method of the intelligent algorithm are used to identify the gestures.Compared with the two methods,it is found that the CNN algorithm can improve the accuracy of gesture recognition by increasing the number of training times.However,it takes a long time to prepare in advance,and the k-cosine curvature method can recognizing the gesture while locate the position of the fingertip,which is more suitable for gesture recognition in real-time video.In order to realize the comfortable gesture non-contact interaction,this paper proposes a handwritten annotation of the distance trigger mechanism,and uses Kalman filtering to track and smooth the fingertip trajectory,so that the user can smoothly and comfortably perform the non-contact interaction of real-time handwritten.Finally,in order to verify the method proposed in this paper,a non-contact demonstration interactive function system was designed.An improved threshold segmentation method was used to segment the gesture,then the k-cosine curvature algorithm was used to recognize the gesture and fingertip detection and tracking.The handwriting presentation function is implemented using the annotation method of the distance trigger mechanism.The final experimental results are good,indicating the validity and rationality of the proposed method.