The present invention provides a system and method for enabling meaningful body-to-body interaction with virtual video-based characters or objects in an interactive imaging environment including: capturing a corpus of video-based interaction data, processing the captured video using a segmentation process that corresponds to the capture setup in order to generate binary video data, labeling the corpus by assigning a description to clips of silhouette video, processing the labeled corpus of silhouette motion data to extract horizontal and vertical projection histograms for each frame of silhouette data, and estimating the motion state automatically from each frame of segmentation data using the processed model. Virtual characters or objects are represented using video captured from video-based motion, thereby creating the illusion of real characters or objects in an interactive imaging experience. Meaningful responses to input human motion are enabled by recognizing patterns of live motion and establishing correspondences between these recognized patterns and the appropriate recorded motion responses.