Characterizing Actions with Local Descriptors Based on Kinematics and Flow Recurrences
A common approach to characterize and recognize human actions consists of detecting interest points within the video volume and describe them locally, followed by a bag-of-words representation. Many of the proposed descriptors are based on the local optic flow, but they may simply summarize the flow in terms of histograms of its orientations. However, potentially interesting and discriminative properties of the optic flow are arguably ignored this way. This work addresses this issue by exploring two new optic flow-based descriptors. One of them consists of kinematic features of spatial variations of the optic flow. The other one captures dynamic patterns of the optic flow in terms of its temporal recurrences. It is experimentally found that these descriptors perform competitively with respect to state-of-the art descriptors. Further elaboration of the proposed descriptors and additional experimentation is required to better assess their potential.
KeywordsAction characterization Interest points Local descriptors Optic flow Kinematic features Recurrences
Unable to display preview. Download preview PDF.
- 2.Agustí, P., Traver, V.J., Marin-Jimenez, M.J., Pla, F.: Exploring alternative spatial and temporal dense representations for action recognition. In: Real, P., Diaz-Pernil, D., Molina-Abril, H., Berciano, A., Kropatsch, W. (eds.) CAIP 2011, Part II. LNCS, vol. 6855, pp. 364–371. Springer, Heidelberg (2011)CrossRefGoogle Scholar
- 4.Bregonzio, M., Xiang, T., Gong, S.: Fusing appearance and distribution information of interest points for action recognition. PR 45(3), 1220–1234 (2012)Google Scholar
- 5.Chakraborty, B., Holte, M.B., Moeslund, T.B., González, J.: Selective spatio-temporal interest points. CVIU 116(3), 396–410 (2012)Google Scholar
- 6.Dollar, P., Rabaud, V., Cottrell, G., Belongie, S.: Behavior recognition via sparse spatio-temporal features. In: International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance, pp. 65–72 (2005)Google Scholar
- 10.Li, R., Zickler, T.: Discriminative virtual views for cross-view action recognition. In: CVPR (2012)Google Scholar
- 13.Schüldt, C., Laptev, I., Caputo, B.: Recognizing human actions: A local SVM approach. In: ICPR, pp. 32–36 (2004)Google Scholar
- 15.Wang, H., Ullah, M.M., Kläser, A., Laptev, I., Schmid, C.: Evaluation of local spatio-temporal features for action recognition. In: BMVC (2009)Google Scholar