Human Motion Data Retrieval Based On Semantic Feature

With the development of motion capture,many large-scale three-dimensional human motion databases are emergenced,and human motion retrieval receives increasing attention.Most existing human motion retrieval techniques are content-based retrieval,these approaches define a set of features representation of motion by analyzing motion data,but features are non semantic and low-level,which make people difficult to understand and lead to "semantic gap",so the results of three-dimensional human motion retrieval are unsatisfactory.To solve these problems above,previous work is analyzed and summarized,semantic-based human motion retrieval is discussed,especially,the semantic gap between low-level features and semantic features.The main work is as follows:According to low-level features and kinesiology,a geometric features coding function and encoding condition are defined,a human motion is filtered into a geometric code stream,which is called semantic features.By incorporating semantic features with human skeleton,a semantic indexing tree is constructed,and feature index is calculated.The hierarchical clustering algorithm is applied to organize feature database.To search logically similar motion,a hierarchical retrieval method is designed based on the representation of motion features,includes retrieving feature index,retrieving semantic feature and retrieving motion data.An efficient human motion retrieval system based on semantic feature adopting hierarchical retrieval method is exploited.Human motion retrieval based on semantic feature could compare motion in large-scale motion database effectively through this study,which contributes towards the reuse of motion data to create new,realistic human animations,and promotes the development of game,entertainment,simulation area.