Understanding Sports Video Using Players Trajectories - Intelligent Video Event Analysis and Understanding - page 142

4.2 Handball Activity Modeling Using HPaSMM

This section describes the HPaSMM used to process squash videos. As illustrated by
Figure 10, it relies on eight upper level states and on the five feature values defined
in the Section 4.1.

Fig. 10 HPaSMM modeling of handball composed of eight upper level states S i . Each state
corresponds to a given phase of activity. A phase of activity is modeled by a five-layer
PaHMM (one layer for each feature vector) and by a GMM modeling the state durations ( sd i
is the state duration associated to state S i ). Upper level layer is surrounded by a red square
and lower level layer is surrounded by green squares. For display reasons, feature vectors
D S GC are denoted by D i GC , and the same holds for feature vectors D S C , , D S i

intramin , D S i intramean

and D S i

intramax .

Upper layer: Activity modeling by semi-Markov chains

Handball activity phases are defined by HPaSMM upper level states S i . We consider
here eight phases of activity , defined along with their numbering as follows:

- “slowly going into attack”: activity phase 1,

- “attack against set-up defense”: activity phase 2,