Retrieval of similar behavior data using Kinect data

In recent years, the reuse of motion capture data has been demanded from a large database system of motion data with the increasing amount of 3D CG contents based on 3D character animation. Therefore, a method to efficiently retrieve motion data is required for the database of motion data. In this paper, we present a method to retrieve similar behavior data in database by using Kinect motion sensor and Dynamic Time Warping (DTW), which is frequently used as an algorithm for measuring similarity between two temporal sequences. The proposed system conducted the normalization of query data captured by Kinect and the calculation of the area of query data to be employed as feature parameter. In addition, DTW measures the similarity between the query data and the motion data in database. We performed the retrieval of similar behavior data by the proposed system, and obtained the retrieval performance of approximately 71%.