Gait Analysis and Human Motion Tracking - Intelligent Video Event Analysis and Understanding - page 65

parameters, represented as a truncated Fourier series of up to three harmonics for
each component of motion. Experiments on synthetic and real data have shown
that the proposed strategy has more accurate and efficient ego-motion estimates and
structural recovery than a comparable method that does not incorporate long-term
motion estimates. The method is robust, and works well in an environment in which
the illumination is relatively static. Constantly moving or changing shadows, caused
for example by moving bushes or clouds covering the sun, would inevitably cause
more features to appear, disappear or shift, and would necessitate an even more ro-
bust and complicated algorithm. This is beyond the current scope and would have
to be included in future work.
Acknowledgments. We thank Iain Wallace for helping generate the computer game scenarios
using the Quake game engine.
References
1. Azarbayejiani, A.J., Galyean, T., Horowitz, B., Pentland, A.: Recursive estimation for
CAD model recovery. In: Proc. of the 2nd CAD-Based Vision Workshop, pp. 90-97.
IEEE Computer Society, Los Alamitos (1994)
2. Beaton, A., Tukey, J.: The fitting of power series, meaning polynormials, illustrated on
band-spectroscopic data. Technometrics 16, 147-185 (1974)
3. Broida, T., Chellappa, R.: Estimation of object motion parameters from noisy images.
IEEE Trans. on Pattern Analysis and Machine Intelligence 8, 90-99 (1986)
4. Cappozzo, A.: Analysis of the linear displacement of the head and trunk during walking
at different speeds. J. Biomechanics 14, 411-425 (1981)
5. Choi, K., Carcassoni, M., Hancock, E.: Recovering facial pose with the EM algorithm.
Pattern Recognition 35(10), 2073-2093 (2002)
6. Coorg, S., Teller, S.: Spherical mosaics with quaternions and dense correlation. Interna-
tional Journal of Computer Vision 37(3), 259-273 (2000)
7. Cross, R., Hancock, E.: Graph matching with a dual-step EM algorithm. IEEE Transac-
tions on Pattern Analysis and Machine Intelligence 20, 1236-1253 (1998)
8. Davison, A., Molton, N., Reid, I., Stasse, O.: MonoSLAM: Real-time single camera
slam. IEEE Trans. Pattern Anal. Mach. Intell. 29(6), 1052-1067 (2007)
9. Demidenko, E., Stukel, T.A.: Efficient estimation of general linear mixed effects models.
J. of Stat. Plan. and Infer. 104, 197-219 (2002)
10. Dempster, A.: Covariance selection. Biometrics 28, 157-175 (1971)
11. Dempster, A., Laird, N., Rubin, D.: Maximal likelihood from incomplete data via the
EM algorithm. J. R. Statist. Soc. B 39, 1-38 (1977)
12. Han, B., Comaniciu, D., Zhu, Y., Davis, L.: Sequential kernel density approximation and
its application to real-time visual tracking. IEEE Trans. Pattern Anal. Mach. Intell. 30(7),
1186-1197 (2008)
13. Hartley, R., Zisserman, A.: Multiple View Geometry in Computer Vision, 2nd edn. Cam-
bridge University Press, Cambridge (2004)
14. Jang, D., Choi, H.: Active models for tracking moving objects. Pattern Recognition 33,
1135-1146 (2000)