[Show abstract][Hide abstract]ABSTRACT: This paper aims at the construction of the system that explains pictures contents considering storyness. The input to this system is information on some pictures and the output of this system is the subjective story-like explanation of pictures contents. A present system consists of a basic contents explanatory part, a connective relationship explanatory part and a story composition part. The basic contents explanatory part explains the behavior or feelings of objects drawn in pictures. The connective relationship explanatory part gives the explanation of the connection between two pictures. The story composition part composes the descriptions obtained from other two parts story-likely. This paper also shows the simulation experiments for the confirmation of the usefulness of the present system.
[Show abstract][Hide abstract]ABSTRACT: This paper aims at the construction of a system explaining some pictures by words consistently. Input to this system is information on some pictures and this system outputs explanation of basic contents of pictures and consistent connective relationships between pictures. The present system consists of a basic content explanatory part and a connective relationship explanatory part. A basic content explanatory part explains the behavior of objects drawn in pictures. A connective relationship explanatory part explains things not drawn in pictures by guessing them from pictures. This part considers consistency of connective relationships. Simulation experiments are performed in order to confirm the usefulness of the present approach. In the experiments the model of the individual subject is constructed and the outputs of the model are evaluated. Experiment results show that good evaluation is obtained.
[Show abstract][Hide abstract]ABSTRACT: This paper describes the system that outputs linguistic expressions of pictures considering consistency of their contents and connections, where pictures are given in any order. Picture information obtained from given pictures is inputted into the system. The system gets descriptions of basic contents of each picture using neural network models, fuzzy reasoning and case-based reasoning. The connective relationships between pictures are also explained using case-based reasoning. If the contradictory relationships between pictures are obtained, they are eliminated from linguistic expressions. The system outputs linguistic expressions including basic contents of pictures and relationships between pictures. Finally, simulation experiments show that the presented system is useful
[Show abstract][Hide abstract]ABSTRACT: This paper aims at constructing a recognition model of emotions from not only real face images but also situations in which human is put. Features values extracted from a real face image by the image processing are regarded as information on facial expressions. Information on a situation is transformed into emotions using questionnaire data about situations. These two pieces of information are regarded as inputs to the recognition model that is composed of seven kinds of neural network models corresponding to six basic emotions and unnatualness. The outputs of the recognition model are numerical values in [0, 1] representing the degrees of emotions or unnaturalness. The questionnaire about the relationship between the combination of facial expressions and situations and the degree of emotion is performed in order to obtain learning data. The recognition models are obtained using questionnaire data. In order to verify the validity of constructed models, models outputs for the combinations of facial expressions and situations that are not used for the construction of the models are evaluated. The evaluation results show that the constructed models recognize happiness and surprise well but that the models feel various emotions for unpleasant ones.