A Proposal of a Multimedia Cooperative Drama Scene Recognition System

Several kinds of studies on color image analysis have been cam'ed out, such as aufornatic k e y words extraction from color images and aulomntic classification of colos image sequence. But, the recognition level of cofor images is not enough high to recognize all things from color images alone. OR the other hand, the recognition of drama scene is necessary Jotef ic ienf handling of color moving media, such as creation o/ a new scene modified jrorn the original scene. In this pnper, a multimedia coopemtn've drama scene recognition system using common concepts is proposed. This sysfem employs so called fusion of the information jmm various medin such as images and that from scenario and sound, in order l o carry out higher recogni!io.m of a given dmmn scene. To show the effrcfiueness c?f this syslem, experiments were carried out using o bod 7 minutes T V dmma scene. According fo the resufts of Ehe experiments, this system could sttccesslull^, mnke n correspondence between images nnd scenario z~sing common concepts, s t ~ c h ns lime, number of people, conversn tion pattern oJtwo people, and dzrmtion lime of the scene. and David M.McKeown combined disparity map and monocular image to estimate threedimensional scene structure[4]. On the other hand, the recognition of drama scene is necessary for efficient handling of color moving media, such as creation of a new scene modified from the original scene. In this paper, n multimedia cooperative drama scene recognition system using cornmon concepts is proposed. This system employs so called fusion of the information from various media such as images and that from scenario and sound, in order to carry out higher recognition of a given drama scene. To show the effectiveness of this system, experiments were carried out using 7 minutes TV drama scene. According to the results of the experiments, this system could successfully make a correspondence lretween images and scenario using common concepts, such as time, number of people, conversation pattern of two people, and duration time of the scene. The details on the multimedia cooperative drama scene recognition system are discussed in section 2. In section 3, the method to make the correspondence between cuts and scene are shown. To show the effectiveness of this system, the results of the experiments are also