In this paper we propose a scripting language that enables
authors to describe and generate emotional eye movement
in virtual agents. The overall system of our research is illustrated
in Fig. 1. The EEMML is based on emotional eye
movement generation framework which picks the parameters
from the AU-Coded facial expression database [1] and realtime
eye movement data (pupil size, blink rate and saccade).
We compute statistics about the involved FAP [4] which contains
the values of eyebrow, eyelid and eyeball that associate
with the eye movement. Finally, a rule based method is
employed to generate emotional eye movement for primary
and intermediate emotions in virtual agents. The realization
of the framework is extensively discussed in paper[2].
The tags are defined to control the agents’ emotional eye
movement. Document Type Definition (DTD) for the tags
is created to define the grammar. The validation module
invokes the Document Type Definition (DTD) file to check
the errors contained in the EEMML script file. The parser
module employs the Simple APIs for XML (SAX) to parse
the EEMML script. The converter module transforms the
parsed EEMML script to FAP values according to the emotional
eye movement generation framework.