Actions as Space-Time Shapes

Abstract

Abstract-Human action in video sequences can be seen as silhouettes of a moving torso and protruding limbs undergoing articulated motion. We regard human actions as three-dimensional shapes induced by the silhouettes in the space-time volume. We adopt a recent approach [14] for analyzing 2D shapes and generalize it to deal with volumetric space-time action shapes. Our method utilizes properties of the solution to the Poisson equation to extract space-time features such as local space-time saliency, action dynamics, shape structure, and orientation. We show that these features are useful for action recognition, detection, and clustering. The method is fast, does not require video alignment, and is applicable in (but not limited to) many scenarios where the background is known. Moreover, we demonstrate the robustness of our method to partial occlusions, nonrigid deformations, significant changes in scale and viewpoint, high irregularities in the performance of an action, and low-quality video.

Continuous Human Action Recognition in Ambient Assisted Living Scenarios

“…Our approach has been validated on the multi-view INRIA XMAS (IXMAS) [20] dataset and the single-view Weizmann [21] dataset. The former provides continuous multi-view sequences of different actions performed by the same actor, whereas the latter provides segmented single-view sequences.…”

Hollywood 3D: What are the Best 3D Features for Action Recognition?

“…Generally speaking interest points are also filtered via non-maxima suppression, to prevent duplicate entries. Some approaches segment the actor, for example using the Kinect's user mask (Gorelick et al 2007;Li et al 2010;Cheng et al 2012). This enables complex "volumetric" descriptions of the actor's body over time (Yang et al 2012;Wang et al 2012;Vieira et al 2012;Oreifej and Liu 2013).…”

Reduced Analytical Dependency Modeling for Classifier Fusion

“…With these notations and adding aand Flower [19] databases, respectively. After that, the results for face recognition using CMU PIE [20] and FERET [21], and human action recognition using Weizmann [22] and KTH [23] databases, are reported in Sections 4.3 and 4.4, respectively. It is important to point out that the main objective of these experiments is to evaluate the performance of different classifier fusion methods, but not state-of-art digit, flower, face and human action recognition algorithms.…”