REAL-TIME CAPTIONING WITH AUTOMATED SPEECH RECOGNITION

描述：该项目将研究在真实的中使用ASR的可行性- 为听障人士提供时间字幕服务。原型系统将 被开发为使用语音阴影的现场广播字幕， 语音被识别并用于产生电视信号的操作员 字幕阴影字幕可以在扬声器适应上执行 识别器具有相对较低的错误率。为了评估该系统， 它的输出将与商业速记字幕的输出进行比较 技术.第一阶段的目标是建立阴影字幕， 一个有竞争力的替代品。在第二阶段，全面 开发的原型将在现场条件下进行评估。长期 项目目标包括生产和商业化负担得起的， 基于PC的语音识别系统，可由 电视节目制作人的一小部分，目前的字幕成本。 拟议商业应用：不可用

DESCRIPTION: The project will study the feasibility of using ASR in real- time captioning services for the hearing impaired. A prototype system will be developed for captioning live broadcasts using speech shadowing by an operator whose speech is then recognized and used to generate television captions. Shadow captioning can be performed on speaker-adapted recognizers which have relatively low error rates. To evaluate the system, its output will be compared to that of commercial stenocaptioning technology. The objective of Phase I is to establish shadow captioning as a competitive alternative to stenocaptioning. In Phase II, a fully developed prototype will be evaluated under field conditions. Long-term project goals include the production and commercialization of affordable, PC-based speech recognition systems that can be purchased and operated by television program producers at a fraction of current captioning costs. PROPOSED COMMERCIAL APPLICATION: NOT AVAILABLE