Moving TV image analysis based on multimedia fusion focusing on extracted common concepts

In some future industrial applications of the multimedia field, moving TV images could give useful information. However, the automatic recognition level of such color moving images achieved by computer is not high enough to understand all things. To overcome this problem, a new method based on multimedia fusion and integration using extracted common concepts is proposed in this paper. In the proposed method, common concepts, such as time and number of people are extracted from multiple data sources such as image, scenario document and sound. Cooperation between various media is carried out by exchanging these common concepts. According to the results of the experiments using a TV drama scene, this system could successfully realize good synchronizations between image, scenario and sound. This system also could realize personal character identification