Multi-Modal Classifier-Fusion for the Recognition of Emotions - Coverbal Synchrony in Human-Machine Interaction - page 80

Figure 4. Example for the classifi cation of arousal in human-computer interaction. The strong
colors represent decisions that were used in the fusion (solid line), whereas the lighter dots
represent rejected samples.
(Color image of this fi gure appears in the color plate section at the end of the topic.)
Figure 5. Multiple classifi er architecture, which is making use of the reject option. The
classifi cation result of each channel has to pass a rejection step, in which decisions with low
confi dences are fi ltered out. The outcome is temporally fused and combined.
classifi cation but also to reconstruct missing values. Finally a classifi er
combination is conducted.
2.2 Base classifiers
In our architectures, linear classifiers, artificial neural networks (e.g.
multilayer perceptrons (MLPs)) and support vector machines (SVMs)