Rule Based Trajectory Segmentation Applied to an HMM-Based Isolated Hand Gesture Recognizer

  • Jounghoon Beh
  • David Han
  • Hanseok Ko
Part of the Communications in Computer and Information Science book series (CCIS, volume 174)


In this paper, we propose a simple but effective method of modeling hand drawn gestures based on their angles and curvature of the trajectories. Each gesture trajectory is composed of a unique series of straight and curved segments. In our Hidden Markov Model (HMM) implementation, these gestures are modeled as connected series of states analogous to series of phonemes in speech recognition. The novelty of the work presented here is the automated process we developed in segmenting gesture trajectories based on a simple set of threshold values in curvature and accumulated curvature angle. In order to represent its angular distribution of each separated states, the von Mises distribution is used. Likelihood based state segmentation was implemented in addition to the threshold based method to ensure that gesture sets are segmented consistently. The proposed method can separate each angular state of training data at the initialization step, thus providing a solution to mitigate ambiguity on initializing HMM. For comparative studies, the proposed automated state segmentation based HMM initialization was considered over the conventional method. Effectiveness of the proposed method is shown as it achieved higher recognition rates in experiments over conventional methods.


Trajectory segmentation hand gesture recognition hidden Markov model HMM initialization 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Starner, T., Weaver, J., Pentland, A.: Real-time American sign language recognition using desk and wearable computer based video. IEEE Trans. PAMI 20(12), 1371–1375 (1998)CrossRefGoogle Scholar
  2. 2.
    Garg, P., Aggarwal, N., Sofat, S.: Vision based hand gesture recognition. World Academy of Science, Engineering and Technology 49 (2009)Google Scholar
  3. 3.
    Rao, G., Yilmaz, A., Shah, M.: View-inveriant representation and recognition of actions. International Journal of Computer Vision 50(2), 203–226 (2002)CrossRefzbMATHGoogle Scholar
  4. 4.
    Gunter, S., Bunke, H.: HMM-based hand written word recognition: on the optimization of the number of states, training iterations and Gaussian components. Pattern Recognition 37, 2069–2079 (2004)CrossRefGoogle Scholar
  5. 5.
    Lee, J.J., Kim, J., Kim, J.H.: Data-driven design of HMM topology for on-line handwriting recognition. In: The Proceedings of 7th International Workshop on Frontiers in Handwriting Recognition, pp. 239–248 (2000)Google Scholar
  6. 6.
    Gunter, S., Bunke, H.: Optimizing the number of states, training iterations and Gaussian s in an HMM-based Handwritten Word Recognizer. In: The Proceedings of the 7th International Conference on Document Analysis and Recognition (2003)Google Scholar
  7. 7.
    Juang, B.-H., Rabiner, L.R.: A probabilistic distance measure for hidden Markov models. AT&T Technical Journal 64(2), 391–408 (1985)MathSciNetCrossRefGoogle Scholar
  8. 8.
    Baum, L.E.: An inequality and associated maximization technique in statistical estimation for probabilistic functions of Markov processes. Inequalities 3(9), 1–8 (1972)MathSciNetGoogle Scholar
  9. 9.
    Fisher, N.I.: Statistical Analysis of Circular Data, Ch. 3.3.6 The von Mises distiribution, pp. 48–56. Cambridge University Press, Cambridge (1993)CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2011

Authors and Affiliations

  • Jounghoon Beh
    • 1
  • David Han
    • 2
  • Hanseok Ko
    • 1
    • 3
  1. 1.Institute for Advanced Computer StudiesUniversity of MarylandCollege ParkUSA
  2. 2.Office of Naval ResearchUSA
  3. 3.School of EEKorea UniversitySeoulKorea

Personalised recommendations