Content Based Querying and Searching for 3D Human Motions

Abstract. From cartoons to medical research world, from the field of arts to scientific visualization, everyone related to these areas have been continuously using 3D motion capture data. Querying on these databases is one of the important tasks. In recent years, almost all the querying interfaces for motion databases are related to tag based search. A lot of manual work is involved in tagging each motion in a database. The focus of this research is on giving the user a freedom of expression by allowing him/her to draw queries and developing a technique to retrieve motions to find an approximate match. In our approach, we have introduced a visual query interface in which a naïve user can draw a query expressed in different human body segments. An indexing algorithm is used to produce the result for a given query format, supported by an efficient pre-processing method for large motion databases.