Linguistic Expression Generation Model of Subjective Content in a Picture

Abstract

As a first step in modeling human intelligent information processing, we describe a model that extracts subjective content from a given picture using objective picture information and generates linguistic expressions. In this model, subjective content is emotions of a human object, the relationship between two objects, and object behavior in a picture. Objective picture information includes object location, size, direction, etc. Model reasoning involves soft computing techniques such as neural networks, fuzzy reasoning, and case-based reasoning. Neural networks recognize human emotions from facial expressions. Fuzzy reasoning infers the degree to which an organism discerns other objects. Case-based reasoning draws object behavior from a picture. This subjective content extracted from a picture is expressed by linguistic expressions using fuzzy sets. A simulation example shows that this model extracts subjective content about an object in a picture. The model\'s effectiveness was confirmed by two types of experiment.