The Improvement of Segmentation and Clustering Algorithm
Based on Velocity and Displacement of Human Motion

Last several years, human motion data is practically used in some domains such as SFX movies, CGs and so on. However, the problem is that it takes many labour work and time to produce motion data. Motion database can help creators to produce motion data with less cost. The database, however, requires content-based retrieval method because it is hard to identify actual motion to map motion keyword to. In this paper, we introduce the method that discovers dependency between contents as association rules. Association rules represent the contents of motion and can be used as visual tags. However, motion data is multi-stream data of time series for body parts, so the amount of data is huge. Then, we impose temporal constraint and make use of the structure of human body. This is because, meaningful dependencies exist: 1) in certain block of time and 2) between body parts groups (i.e. arms and legs). We also apply the algorithm known in data mining domain to reduce the cost of discovery task effectively. We had experiments to discover association rules from multi-stream of motion. The cost of discovery task was largely reduced.