著者
加藤 士雄 井野 秀一 永井 謙芝 渡邉 括行 堀 耕太郎 似鳥 寧信 高杉 弘子 服部 裕之 伊福部 達
出版者
Human Interface Society
雑誌
ヒューマンインタフェース学会研究報告集 : human interface (ISSN:13447270)
巻号頁・発行日
vol.4, no.4, pp.65-70, 2002-11-22
被引用文献数
11

In order to enable the profoundly hearing impaired to recognize speeches talked at international conferences, we have designed an automatic captioning system using a speech recognition technology. This system was utilized at a Disabled Peoples' International conference (DPI) held in Sapporo this October. In this system, some speakers (shadowing speakers) including broadcast announcers repeated the same voices as the conference talker's or English interpreter's voices, while the speech recognition device converted the shadowing speaker's voices into JApanese letters or English letters. Where, before using the captioning system, the shadowing speakers have had completed an enrolling their voices into a voice dictionary of the speech recognition device. The conference talker's face images as well as the caption letters were displayed onto a large screen. The conference talker's images were delayed in order to be synchronized with the subtitles. Furthermore, incorrectly recognized words were corrected by humans before displaying them. We have evaluated the captioning system from a viewpoint of a comprehension of the subtitles, the optimal time delay of the talker's image, and a system usability. This report describes the evaluation results which were obtained from the above international conference.