| Shape Time Discriminative Classification of Video Objects (2008) | |||||||||||||||
Abstract | |||||||||||||||
| We propose a discriminative approach to non-rigid video objects classification. Our goal is to recognize actions of the objects that appear in a video sequence, based on its shape time dynamics. This is achieved by exploiting the temporal action correlations under large-margin structured classification framework. Further, it leads to an algorithm that can naturally utilize non-vectorial shape representation and matching techniques, which are difficult to incorporate into the commonly used hidden Markov models (HMMs). The proposed approach is verified on indoor video sequences for action classifications of human subjects. 1 | |||||||||||||||
Publication details | |||||||||||||||
| |||||||||||||||