Semiotic modeling of human behaviors interacting with the artifacts using recurrent neural networks

For realizing an interface system that can grasp the intentions of the operator lying behind the observed interactions with the system, some intelligent mechanism for detecting the contextual features of serial data is needed. We introduce a recurrent neural network (i.e., Elman network) for this purpose. After analyzing its generalization capability over a collection of artificial serial data having different contextual features, then we apply this network for extracting the segmentation boundaries of the observed human-system interactions