Saturday , June 23 2018

Automatic Parameter Configuration: A Case Study on a
Rehabilitation Oriented Human Limb Tracking Algorithm


1 Facultad de Matemática y Computación,
Universidad de la Habana, Cuba
2 Unitat de Gràfics, Visió i Intelligència Artificial,
Departament de Ciències Matemàtiques i Informàtica,
Universitat de les Illes Balears, Spain

Abstract: The choice of the parameters has strong influence on the quality of the results obtained by the application of algorithms. Most researchers tend to select the values of their parameters in long and tedious trial and error approaches. Although, some methods have been developed for automatic parameter selection, they have not been widely used in the computer vision area. This paper presents the design of a general purpose framework for automatic parameter selection through a case study: a human limb tracking algorithm developed for applications that will be used in rehabilitation scenarios with low cost equipment. The tracking algorithm first detects the limb by using a skin segmentation approach, then the position of an idealized limb model is updated using Simulated Annealing. The framework for automatic parameter selection treats each parameter from the tracking algorithm according to its domain and uses a modified version of Harmony Search Optimization algorithm that includes a dominance criterion. The obtained results are presented as well and show that selected parameters behave well for the case of study.

Keywords: Automatic parameter selection; Harmony Search Optimization; Simulated Annealing; Human limb tracking.

>>Full text
Oscar L. VERA-PÉREZ, Alejandro MESEJO-CHIONG, Antoni JAUME-I-CAPÓ, Manuel GONZÁLEZ-HIDALGO, Automatic Parameter Configuration: A Case Study on a Rehabilitation Oriented Human Limb Tracking Algorithm, Studies in Informatics and Control, ISSN 1220-1766, vol. 23 (1), pp. 87-96, 2014.

  1. Introduction

The choice of the parameters has strong influence on the behavior of algorithms. In most cases the stability of parametrized algorithms and the quality of obtained results are heavily affected by parameters. Finding the right parameter values, when solving a specific kind of problem with certain algorithm, is a very difficult task and even harder in the presence of a non-deterministic algorithm such as metaheuristics.

Some authors state that most of the time spent in the test and design of some metaheuristic algorithm is devoted to find good parameter configurations that lead to acceptable results. Many researchers tend to fine-tune the parameters of their algorithms using some long and tedious trial and error approaches or they simply take previously reported parameter values without any analysis on whether those parameters are suitable or not for their specific cases. Just a few make correct use of experimental designs for the parameter selection matter [1].

A recently developed algorithm for human limb tracking [22] based on the Simulated Annealing metaheuristic [14], immersed of course in the computer vision discipline, made use of a general purpose framework for automated parameter selection with good results. We designed this algorithm to be used in rehabilitation sessions of patients with physical disabilities, in particular persons that have little mobility in their shoulders. In these sessions, the patients should perform repetitive movements with their limbs in an indoor environment with controlled illumination and background. Also, the equipment for practical implementation is expected to be low cost, probably only a web cam and a regular PC. The tracking algorithm can facilitate progress analysis of the patients’ condition and the implementation of new rehabilitation systems [12], such as serious games.

As it is very unlikely that it can be found a parameter configuration that works well for every possible scenario of application, the used general purpose framework obtains good parameters in a supervised learning way using a representative set of problem instances that resemble the conditions in which the tracking algorithm should work. According to a dominance definition, non-dominated parameter vectors are selected using a modified version of Harmony Search Optimization algorithm [26] that includes those dominance notions and takes advantage of the domain of definition for every parameter. The complete selection process may be viewed as an automatic experimentation phase that aids to perform deeper analysis of the algorithms being configured, in this case, the proposed tracking algorithm optimizing the relationship between the image processing, the capture device, and the rehabilitation motions.

In the present paper, we expanded previous work [22] by presenting in detail the specific tracking problem definition, the designed tracking algorithm and the automated parameter selection strategy used in order to get good results on the scenario of application. First, we give a suitable tracking problem definition. Next, a review of related work is presented. After that, the tracking algorithm is explained and the last sections are devoted to relate the details of automatic parameter selection and the analysis of the obtained experimental results. Finally some conclusions of the presented work are discussed.


  1. Adenso-Diaz, B., M. Laguna, Fine-tuning of Algorithms using Fractional Experimental Designs and Local Search. Operations Research, vol. 54(1), 2006,   pp. 99-114.
  2. Battiti, R., M. Brunato, F. Mascia, Reactive Search and Intelligent Optimization, Op. Res. / Comp. Sc. Interfaces, vol. 45, 2009.
  3. Birattari, M., Z. Yuan, P. Balaprakash, T. Stutzle, F-race and Iterated f-race: An Overview. in T. Bartz-Beielstein, M. Chiarandini, L. Paquete, and M. Preuss, (Eds.): Experimental methods for the analysis of optimization algorithms, Springer-Verlag New York Inc, 2010, pp. 311-336.
  4. Brumby, S. P., N. R. Harvey, S. Perkins, R. B. Porter, J. J. Szymanski, J. Theiler, J. J. Bloch, A Genetic Algorithm for Combining New and Existing Image Processing Tools for Multispectral Imagery. in: S. S. Shen and M. R. Descour, (Eds.): Algorithms for Multispectral, Hyperspectral, and Ultraspectral Imagery VI. Proceedings of SPIE, Vol. 4049, 2000, pp. 480-490.
  5. Cotta, C., M. Sevaux, K. Sorensen, Adaptive and Multilevel Metaheuristics, Studies in Computational Intelligence, vol. 136 Springer Verlag, 2008.
  6. Coy, S. P., B. L. Golden, G. C. E. A. Runger, G. C. Wasil, Using Experimental Design to Find Effective Parameter Settings for Heuristics. Journal of Heuristics, vol. 7(1), 2001, pp. 77-97.
  7. Gall, J., B. Rosenhahn, T. Brox, H. P. Seidel, Optimization and Filtering for Human Motion Capture. International Journal of Computer Vision, vol. 87(1-2), 2010, pp. 75-92.
  8. Gall, J., B. Rosenhahn, H. P. Seidel, An Introduction to Interacting Simulated Annealing. In: R. Klette, D. Metaxas, and B. Rosenhahn, (Eds.):Human Motion – Understanding, Modeling, Capture and Animation, Computational Imaging, Springer, Heidelberg. vol. 36, 2008, pp. 319-345.
  9. Geem, Z. W., Music Inspired Harmony Search Algorithm: Theory and Applications, Studies in Computational Intelligence. Springer Verlag, vol. 191, 2009.
  10. GEEM, Z. W., Recent Advances in Harmony Search Algorithm, Studies in Computational Intelligence. Springer Verlag, vol. 270, 2010.
  11. Hutter, F., H. H. Hoos, K. Leyton-Brown, T. Stutzle, Paramils: an Automatic Algorithm Configuration Framework. Journal of Artificial Intelligence Research, vol. 36(1), 2009, pp. 267-306.
  12. Jaume-i-CapÓ, A., P. Martinez-Bueso, B. Moya-Alcover, , J. Varona, Interactive Rehabilitation System for Improvement of Balance Therapies in People With Cerebral Palsy. Neural Systems and Rehabilitation Engineering, IEEE Transactions on. DOI: 10.1109/TNSRE.2013.2279155.
  13. Jaume-i-Capó, A., J. Varona, M. Gonzalez-Hidalgo, F. J. Perales, Adding Image Constraints to Inverse Kinematics for Human Motion Capture. EURASIP J. Adv. Signal Process, Vol. 4, 2010, Art. 4. doi:10.1155/2010/142354.
  14. Kirkpatrick, S., Jr. C. D. Gelatt, M. P. Vecchi, Optimization by Simulated Annealing. Science, Vol. 220, No. 4598, 1983, pp. 671-680.
  15. Lee, G., V. Bulitko, I. Levner, Automated Selection of Vision Operator Libraries with Evolutionary Algorithms. In: Proceedings of the 2004 IEEE Congress on Evolutionary Computation (CEC 2004), IEEE Press, vol. 1, 2004, pp. 1127-1134.
  16. Luo, Y., H. Yang, Z. Hu, Human Limb Motion Real-time Tracking based on Camshift for Intelligent Rehabilitation System. IEEE International Conference on Robotics and Biomimetics (ROBIO), 2009, pp. 343-348.
  17. Minton, S., An Analytic Learning System for Specializing Heuristics. Proceedings of the 13th International Joint Conference on Artificial intelligence, IJCAI’93, vol. 2, 1993, p. 922.
  18. Moeslund, T. B., E. Granum, A Survey of Computer Vision-based Human Motion Capture. Computer Vision and Image Understanding, vol. 81(3), 2001, pp. 231-268.
  19. Moeslund, T. B., A. Hilton, V. Kruger, A Survey of Advances in Vision-based Human Motion Capture and Analysis. Computer Vision and Image Understanding, vol. 104(2), 2006, pp. 90-126.
  20. Poppe, R., Vision-based Human Motion Analysis: An Overview. Computer Vision and Image Understanding, vol. 108(1-2), 2007, pp. 4-18.
  21. Varona, J., A. Jaume-i-Capó, J. Gonzalez, F. J. Perales, Toward Natural Interaction through Visual Recognition of Body Gestures in Real-time. Interacting with Computers, vol. 21(1-2), 2009, pp. 3-10.
  22. Vera-Pérez, O. L., A. Mesejo-Chiong, A. Jaume-i-Capó, M. González-Hidalgo, Human Limb Tracking using Simulated Annealing and Skin Features. Actas del VIII congreso espanol sobre metaheuristicas, algoritmos evolutivos y bioinspirados (in Spanish).
  23. Osorio, R., S. García, M. Peña, I. Lopez-Juarez, G. Lefranc, Movement and Color Detection of a Dynamic Object An Application to a Mobile Robot. Studies in Informatics and Control, 2012, vol. 21, no 1, p. 34.
  24. TANGOUR, F., P. BORNE, Presentation of Some Metaheuristics for the Optimization of Complex Systems. Studies in Informatics and Control, 2008, vol. 17, no 2, p. 169.
  25. Vezhnevets, V., V. Sazonov, A. Andreeva, A Survey on Pixel-based Skin Color Detection Techniques. Proceedings of the GraphiCon 2003, Moscow, Russia, vol. 3, 2003, pp. 85-92.
  26. Wang, C. M., Y. F. Huang, Self-adaptive Harmony Search Algorithm for Optimization. Expert Systems       with Applications, vol. 37(4), 2010,       pp. 2826-2837.
  27. Zhou, H., H. Hu, Human Motion Tracking for Rehabilitation–A Survey. Biomedical Signal Processing and Control, vol. 3(1), 2008, pp. 1-18.