| ECVision indexed and annotated bibliography of cognitive computer vision publications This bibliography was created by Hilary Buxton and Benoit Gaillard, University of Sussex, as part of ECVision Specific Action 8-1 The complete text version of this BibTeX file is available here: ECVision_bibliography.bib
Xiaohui Liu and Chin-Seng Chua
ABSTRACT
A new approach of modeling/recognizing multi-agent activities from image sequences is presented. In recent years, Hidden Markov Models (HMMs) have been widely used to recognize activity units ranging from individual gestures to multi-people interactions. However, traditional HMMs meet many problems when the number of agents increases in the scene. One significant reason for this inability is the fact that HMMs require their `observations' to be of fixed length and order. Unlike conventional HMMs, a new algorithm to model multi-agent activities is proposed. This has two sub-processes: one for modelling the activity based on decomposed observations and the other for recording the `role' information of each agent in the activity. This new algorithm allows changing of the observations' length, and does not require initial agent assignment. The experimental results show that this algorithm is also robust when the agents' information is only partially represented |
|
| Site generated on Friday, 06 January 2006
|