This structure enables the BN inference to take into account shot-class association related information for every shot s i, as well as for all its neighboring shots that lie within a certain time window, for deciding upon the class that is eventually associated with shot s i.
It has been regarded that arg max j ( h i j m ) indicates the class e j that is associated with shot s i.
The estimated posterior probability, denoted by h i j a = P ( e j ∣ O S i a ), indicates this time the degree of confidence with which class e j is associated with shot s i based solely on audio information.
This probability, which represents the observation sequence's fitness to the particular HMM, indicates the degree of confidence with which class e j is associated with shot s i based on motion information.
At the evaluation stage each HMM returns a posterior probability, which is denoted by h i j c = P ( e j ∣ O S i c ) and indicates the degree of confidence with which class e j is associated with shot s i.
In all cases, it has been considered that arg max j ( h i j a ), arg max j ( h i j c ), arg max j ( h i j m ) and arg max j ( h i j f ) indicate the class e j that is associated with shot s i after every respective algorithmic step.
Each of these probabilities indicates the degree of confidence, denoted by h i j f, with which class e j is associated with shot s i. Dynamic Bayesian Networks (DBNs), and in particular HMMs, have been widely used in semantic video analysis tasks due to their suitability for modeling pattern recognition problems that exhibit an inherent temporality (Section 2.1).
During a botched convenience store robbery when he was 19, McIntyre's associate shot a clerk with a.357 magnum twice in the chest at point-blank range all over $20. McIntyre was the getaway driver.
During a botched convenience-store robbery when he was 19, McIntyre's associate shot a clerk with a.357 magnum twice in the chest at point-blank range all for $20. McIntyre was the getaway driver.
