Abstract
The aim of this paper is to compose music that is suitable for the emotion extracted from a visual medium like a picture story book. The values of emotion in the picture are inferred from its hue and color brightness, and these in the story are calculated by matching between the words and the sentiment classification dictionary. Music is automatically composed by genetic algorithms with the fitness function based on the extracted emotion values. First some pieces of music with a bar are composed. By using them as a gene, some pieces of music with 8 bars are composed. Subjects evaluated the composed music with cheerful and gloomy image, the result of which shows the validity of the fitness function.
