Interpolation and extrapolation of motion capture data

The author proposes a computer-graphic animation tool through which the relationship between human motions and subjective impressions can be quantitatively examined. In the system, the three-dimensional human motion was measured with a wireless magnetic motion capture system with a pair of data gloves. The time-series changes of each joint angle were estimated through the measured data, using a method where each center of the joint angles were estimated based on the stochastic behaviors of the sensors whose each temporal average is zero, and then, each joint angle was calculated with position vectors from neighboring two sensors and a joint center-point estimated. The behaviors of joint angles were analyzed in the frequency domain. Finally, joint motions were interpolated or extrapolated in the frequency domain between two patterns of dance motions selected at random, and a new motion was synthesized with the inverse Fourier transform and displayed as a computer-graphic animation in the time domain. As the results, our computer-graphic animation tool can harmoniously interpolate and extrapolate various pairs of approximately periodic dance motions with high accuracy.