동영상 대사 자막을 발화자 주변에 자동 배치하는 시스템의 개발

Abstract

Conventional unedited subtitles of the video are placed in a fixed location of screen, which may cause distraction and visual fatigue to viewers. According to previous studies, it has been proven that the subtitling system placing the subtitle near the speaker may improve viewer’s convenience and quality of visual understanding. We developed an automatic video-subtitle placement system which automatically detected the speaker on the screen and placed the active subtitles near the speaker without disturbing the main figures. Pre-existing speech-to-text, active speaker detection and face detection technology were used to extract the text from the speaker on the screen, and we developed a new algorithm to select the optimal space around the speaker to put the caption of the lines on the video screen. The developed program was tested for the accurate speaker detection and placement of subtitles using several types of video genre. The performance of program was satisfactory showing overall 98.25% of accurate subtitle placement rate. Our algorithm for the automatic subtitle placement system may be used to supplement the shortcomings of conventional subtitle system.