A Non-Document Text and Display Reader for Visually Impaired Persons

项目描述（由申请人提供）：本项目的目标是开发一种基于标准照相手机的计算机视觉系统，使盲人和视障人士能够阅读设备显示和类似形式的非文件视觉信息。这种能力在使用诸如微波炉和DVD播放机等日常用具以及进行许多日常活动（如数纸币）时越来越必要。传统的文本读取系统，如光学字符识别（OCR），目前无法访问这些信息，OCR是用于读取打印文档的。我们提出的软件运行在标准的、现成的照相手机上，并使用计算机视觉算法来分析用户拍摄的图像，检测和读取每个图像中的文本，然后使用合成语音大声朗读。初步的可行性研究表明，目前的移动电话很容易超过这些任务所需的最低处理能力。最初，该软件将读出三种类型的符号：LED/LCD家电显示器、产品或用户定义的条形码以及纸币面额。最终，这些功能将与其他在单独资助下开发的功能集成在一起，例如阅读广泛的印刷文本（包括标志）、识别物体、分析照片和图形等，所有这些功能都可以作为免费或低成本的软件下载给任何手机用户。我们的具体目标是：(1)收集盲人和视障人士拍摄的各种LED/LCD家电显示屏、条形码和美国纸币的真实图像数据库；(2)开发算法对图像进行处理，提取所需信息；(3)在照相手机上实现算法；(4)进行用户测试，建立设计参数，优化人机界面。公共卫生相关性：对于盲人和视障人士而言，就业、经济自给自足和独立的最严重障碍之一是无法充分利用家庭、工作场所、学校或大学中越来越多的带有LED/LCD视觉显示的设备和器具，以及迄今为止康复技术尚未解决的其他类型的文字和符号信息。拟议的研究将产生一种辅助技术系统（用户的成本为零或最低），为大约1000万有严重视力障碍或失明的美国人提供更多的访问这些显示和非文档文本信息的机会。

DESCRIPTION (provided by applicant): The goal of this project is to develop a computer vision system based on standard camera cell phones to give blind and visually impaired persons the ability to read appliance displays and similar forms of non-document visual information. This ability is increasingly necessary to use everyday appliances such as microwave ovens and DVD players, and to perform many daily activities such as counting paper money. No access to this information is currently afforded by conventional text reading systems such as optical character recognition (OCR), which is intended for reading printed documents. Our proposed software runs on a standard, off-the- shelf camera phone and uses computer vision algorithms to analyze images taken by the user, to detect and read the text within each image, and to then read it aloud using synthesized speech. Preliminary feasibility studies indicate that current cellular phones easily exceed the minimum processing power required for these tasks. Initially, the software will read out three categories of symbols: LED/LCD appliance displays, product or user-defined barcodes, and denominations of paper money. Ultimately these functions will be integrated with other capabilities being developed under separate funding, such as reading a broad range of printed text (including signs), recognizing objects, and analyzing photographs and graphics, etc., all available as free or low-cost software downloads for any cell phone user. Our specific goals are to (1) gather a database of real images taken by blind and visually impaired persons of a variety of LED/LCD appliance displays, barcodes and US paper currency; (2) develop algorithms to process the images and extract the desired information; (3) implement the algorithms on a camera phone; and (4) conduct user testing to establish design parameters and optimize the human interface. PUBLIC HEALTH RELEVANCE: For blind and visually impaired persons, one of the most serious barriers to employment, economic self sufficiency and independence is insufficient access to the ever-increasing variety of devices and appliances in the home, workplace, school or university that incorporate visual LED/LCD displays, and to other types of text and symbolic information hitherto unaddressed by rehabilitation technology. The proposed research would result in an assistive technology system (with zero or minimal cost to users) to provide increased access to such display and non- document text information for the approximately 10 million Americans with significant vision impairments or blindness.