VIDEO PROCESSING OF SPEECH FOR HEARING-IMPAIRED PERSONS

实验将使用现代信号处理技术进行 识别和测量感知上重要的技术 用于语音阅读(唇读)的视频信号的特性 包括听力正常的人和听力受损的人。的影响 空间滤波以及常见的视频降质在 朗读演讲稿的表现将被调查和索引 类似于清晰度指数将被开发用于 通过身体测量预测朗读演讲稿的表现。 将开发用于合成视频语音信号的方法并 在感知研究中使用，旨在识别物理 在演讲稿阅读中非常重要的信号特征。 辅助性视觉线索的研究进展 通过朗读演讲进行的交流将使用 合成视频语音信号。在以下位置获得的数据 组件项目将被集成，以开发一个 语段层面的朗读定量模型。

Experiments will be performed using modern signal processing techniques to identify and measure the perceptually important characteristics of video signals for speechreading (lipreading) by both normal-hearing and hearing-impaired persons. The effects of spatial filtering as well as that of common video degradations on speechreading performance will be investigated and an index analogous to the Articulation Index will be developed for predicting speechreading performance from physical measurements. Methods for synthesizing video speech signals will be developed and used in perceptual studies designed to identify the physical characteristics of the signal that are important in speechreading. The development of supplementary visual cues for improved communication by speechreading will be investigated using synthesized video speech signals. The data obtained in the component project will be integrated in order to develop a quantitative model of speechreading at the segmental level.