Office Software Control System for Kinect

This paper studies the recognition of body gesture to achieve human-computer interaction, which defined by physical methods and algorithms matching body gesture to achieve recognition. Physical definition mainly for skeletal point of the relative coordinates define the size of the operation, the matching algorithm used quaternion main points of the characteristic features of bone extraction operation, with better flexible, fast, accurate and robust. We need to establish the action database before the algorithm for matching, We first establish a model library by the extraction of BVH files, and then we should get the 3D position of 15 body joints based on quaternion features, and its characteristic normalized, and make its characteristic normalized to establish a model database, and then use DTW (dynamic time warping) algorithm matches the characteristics of the action under test data and the data in the database, which the nearest distance is what we want to detect motion. Interactive system can be used for office software control, to make teaching more intelligent.