The Study On Recognition Method Of Sign Language Gesture Based On Kinect

With the widespread popularity of electronic products as sign language gesture acquisition devices,it is becoming more and more urgent to study the more and more accurate sign language gesture recognition algorithms.The emergence of the Kinect device not only opens up a new way for the natural human-computer interaction,but also provides reference for the recognition of the sign language gesture.Therefore,in this paper,based on the Kinect,the study of sign language gesture recognition,the specific work is summarized as follows:(1)The method of dividing a sign language gesture is summed up.Firstly,by studying the analysis of "Chinese sign language",the shape of sign language gestures is composed of three characteristics:finger number,finger combination and finger shape;Then,61 basic sign language gestures were summarized.Finally,the classification of basic sign language gestures is given.(2)This paper presents a method for extracting the contour of sign language gestures based on Kinect.In the first place,a sign language gesture model with deep information was acquired through Kinect,using the threshold method to extract the sign language gestures.Then,using the skin color feature,the extracted gesture was optimized to eliminate the residual background information.Based on the region density,the maximal connected region is obtained,and the center of mass of the gesture is extracted.(3)Given a method of recognizing sign language gestures based on joint distance and curvature features.First,calculate the gesture distance feature,find a circle with the center of mass as the center of the whole palm,and then use this circle to divide the gesture contour area into the finger area,the palm area,and the arm area,Finally,using the gesture center of mass as the origin,the finger pixels of the entire finger area are scanned to obtain the distance feature of the outline of the sign language gesture.Secondly,calculate the curvature of the sign language gesture,calibrate an initial pixel,draw a circle around this pixel,and then calculate the area of the circle of the pixel point and the area of the gesture area in the circle on each contour line in sequence along the contour line.The curvature of the gesture contour is calculated by the ratio of the gesture contour falling on the circled portion to the circle area.Finally,these two feature vectors are used as the input of the neural network to complete the final recognition of the sign language gestures.The experimental results show that using the sign language gesture recognition method given in this paper can effectively recognize 61 basic sign language gestures.