A Quantitative Methodology to Evaluate Motion-Based Animation Techniques Gutemberg Guerra-Filho, George Raphael, and Venkat Devarajan University of Texas at Arlington, Arlington, TX USA 76019 guerra@cse.uta.edu, george.thekkanathraphael@mavs.uta.edu, venkat@uta.edu Abstract. We present a novel methodology to quantitatively evaluate the synthesized motion generated by a motion-based animation tech- nique. Our quantitative evaluation methodology provides a measure of how well each algorithm synthesizes motion based on their rotational and translational similarities to the ground truth in a motion database. To demonstrate the effectiveness of our methodology, we focus on tech- niques that combine different motions into a single spliced action where individual motions are performed simultaneously. We implement three splicing algorithms to perform a comparison study based on our quan- titative evaluation methodology. The splicing algorithms considered are spatial body alignment, segmentation-based, and na¨ ıve DOF replace- ment. The spatial body alignment adapts the spliced motion according to this joint correlation and, consequently, performs best under our eval- uation methodology. Keywords: quantitative evaluation, motion-based animation, motion splicing. 1 Introduction In general, the correctness of an algorithm for motion synthesis or analysis is mostly assessed visually. This is very subjective in nature and does not provide a uniform method of evaluation. On the other hand, the performance of any algorithm can be quantitatively evaluated when they are tested with precise data providing sampling of all motion variations in a principled controlled fashion. In this paper, we introduce a novel technique to quantitatively evaluate and compare synthesized motions against actual captured motions. Our quantitative evaluation methodology provides a measure of how well each algorithm synthe- sizes motion based on their rotational and translational similarities to the ground truth in a motion database. Initially, the synthesize motion is time-aligned to the ground truth motion. The time-displacement for each DOF is used to pro- duce a normalized spliced motion. This normalized motion is compared to the ground truth motion by computing a similarity distance based on translational and rotational data at three orders of derivatives. In our evaluation methodology, the input data for the evaluated techniques and the corresponding ground truth output data is obtained from the Human J.M. Allbeck and P. Faloutsos (Eds.): MIG 2011, LNCS 7060, pp. 378–389, 2011. c  Springer-Verlag Berlin Heidelberg 2011