Multimedia indexing is about developing techniques allowing
people to effectively find media. Content-based methods
become necessary when dealing with large databases as people
cannot possibly annotate all available content. Emotions
are intrinsic in human beings and are known to be very important
for natural interactions, decision making, memory,
and many other cognitive functions. Current technologies
allows exploring the emotional space by mean of content-
based analysis of audio and video, but also thanks to other
modalities such as the human physiology.
In this paper, we present the latest development in the
emotion recognition part of SAMMI [18] by mean of an
extensive study on feature selection and the application of
many of the principles we have presented in [17] and [15].
Then, we present the concepts of output thresholding, inverse
thresholding, and profiling which we used for improving
the results of the recognition. Finally, we present a study
on the robustness to rotations and zoom of our feature point
tracking system.
Our experiments on the six prototypical emotions by Ekman
and Friesen presented in the eNTERFACE'05 database
result in as much as 75% correct recognition rate.