Development of Wearable Vision Assistive Device with Scene Text Recognition Feature

我们开发了实现穿戴式视觉辅助设备所需的一些基本技术，这些设备可以自动检测环境中的纸质文件、招牌等上的文本，通过光学字符识别和语音合成将其翻译成语音，并将内容呈现给视力受损的用户。这些技术包括快速精确的自动文本区域检测和跟踪，用于实时字符识别的快速模式匹配，以及使用声音信号的文本定位指导。我们还开发了一个原型系统，并获得了可用性和功能需要改进的知识。

We have developed some building-block technologies required for realizing wearable vision assistive devices that automatically detect text on paper documents, signboards, etc. in the environment, translate it into speech by Optical Character Recognition and Speech Synthesis, and present the contents to the visually-impaired users. The technologies include fast and precise automatic text region detection and tracking, fast pattern matching for real-time character recognition, and text location guidance using sound signals. We have also developed a prototype system, and obtained the knowledge about the usability and functions need to be improved.