Generic Human Action Recognition from a Single Example

We present a novel human action recognition method based on space-time locally adaptive regression kernels and the matrix cosine similarity measure. The proposed method operates using a single example (e.g., short video clip) of an action of interest to find similar matches. It does not require prior knowledge (learning) about actions being sought; and does not require foreground/background segmentation, or any motion estimation or tracking. Our method is based on the computation of the so-called local steering kernels as space-time descriptors from a query video, which measure the likeness of a voxel to its surroundings. Salient features are extracted from said descriptors and compared against analogous features from the target video. This comparison is done using a matrix generalization of the cosine similarity measure. The algorithm yields a scalar resemblance volume with each voxel here, indicating the likelihood of similarity between the query video and all cubes in the target video. By employing nonparametric significance tests and non-maxima suppression, we detect the presence and location of actions similar to the given query video. High performance is demonstrated on the challenging set of action data (Shechtman and Irani 2007b) indicating successful detection of actions in the presence of fast motion, different contexts and even when multiple complex actions occur simultaneously within the field of view of the camera. Further experiments on the Weizmann dataset (Gorelick et al. 2007) and the KTH dataset (Schuldt et al. 2004) for action categorization task demonstrate that the proposed method achieves improvement over other (state-of-the-art) algorithms.