VEML) to annotate instances of the events described in VERL [2]. Another representation technique base on hierar-This paper presents a novel approach for video event repre-chical CASE representation was proposed by M Shah et al. sentation and recognition of multi agent interactions. The in [3] and then enhanced by [4] . proposed approach integrates behavior modeling techniques The dynamic nature of video clips always requires robust based on Generalized Stochastic Petri Nets (GSPN) and

more »

... ing technique that can efficiently treat the uncertainty troduces Petri net marking analysis for better scene under-of the video scenes. Therefore, the Bayesian Networks [5] [6] [7] standing. The GSPN model provides remarkable flexibility in and various HMMs [8-10] have been widely used in the area representation of time dependent activities which usually co-of video event recognition. KMurphy introduced the Dyexist with logical, spatial and temporal relations in real life namic Bayesian Networks (DBN) which generalizes HMMs scenes. The nature of Petri net concept allows efficient model-by improving the state space representation [11] . ing of the complex sequential and simultaneous activities but YIvanov and A.Bobick proposed the stochastic parsing disregards the global scope of a given model. The proposed approach [12] that combines the Coupled HMM (CHMM) for