Abstract
In this thesis, a content-based HUman MOtion Retrieval system, HUMOR, is presented. HUMOR provides multiple convenient and intuitive interfaces for interaction, together with effective indexing and matching algorithms for retrieval. In HUMOR, users can choose convenient input modes, including text, stickman, images, and motion clips, to specify their queries. Later, they can observe retrieval results via graphics images or animation video. Besides, a novel retrieval approach, including indexing and matching, is devised to facilitate the search for human motion. In indexing, we introduce an affine invariant posture representation and propose a SOM-based index map according to the distribution of the raw data. To avoid the curse of dimensionality, the high-dimension feature space of the whole body is decomposed into the direct sum of low-dimension feature spaces of skeletal segments. In matching, the start frame and the end frame of the user query, are used to find candidate clips from the given motion collection. Then the similarity between the user query and each candidate clip are computed by using a dynamic time warping algorithm. The usability of the multi-modal user interface and the effectiveness of proposed retrieval approach are demonstrated.