The Research And Application Of Gesture Recognition Based On Depth Image

Gesture recognition technology based on depth image has replace other conventional gesture recognition technology by the development of depth sensor, with depth image the segment and tracking process of the gesture recognition become easier which both are important step for the success of gesture recognition especially the dynamic gesture recognition. So this paper, first, design a fast and effective segment method based on the depth image, and design outstanding biological characteristics to simple the process of feature extraction and to reduce the dimension of the feature to express the gesture for reducing the processed data. Last using a kernel improving the sparse representation to realize dynamic gesture recognition to enhance the recognition accuracy and rate.The main design method following as:1. Using the Kinect senor to obtain the depth image stream and color image stream of dynamic gesture.2. Combining the color segmentation, depth threshold, skeletal tracking technology to segment the gesture with complex background and overlapping objects.3. Designing palm position, finger grooves number, fingertip number, rotate direction four biological characteristics, using minimum inscribed circle, convex defects, fingertip angle, rotation angle method to extract the four feature.4. Combining kernel with sparse representation to recognize dynamic gesture. Realize the dynamic gesture machine learning with different time series, through learning the sparse coding to build the gesture model and then find the optimal solution of linear equation to reduce the recognition time.