基于图像描述和视觉问答的智能盲人辅助系统

为了解决盲人无法通过眼睛感知周围环境和盲人辅助设备无法提供问答功能的问题,设计了一种智能盲人辅助系统,包含硬件和软件两部分。首先,分别对摄像头、麦克风、全球定位系统（GlobalPositioning System,GPS）模块采集信息进行关键帧抽取、语音转文字、关键数据抽取获得对应的图片、文本、位置信息;其次,核心服务软件利用上述信息进行功能的切换,并将对应图片-文本数据、图片数据、文本-位置数据送入图片描述（Image Caption）模型和视觉问答（Visual Question Answering,VQA）模型处理,得到对应回答信息;最后,语音播报模块利用文字转语音技术将回答信息转成语音信息并反馈。测试结果表明,提出的系统能针对盲人具体的语音问题进行回答,实现特定场景的问答和描述。     

一个基于语音识别的盲人上网辅助系统的设计

针对目前一部分残障人士(如盲人等)面临的上网困难问题,本文通过语音识别和文本-语音转换两个部分的设计及运用,设计一个基于语音识别的盲人上网辅助系统。系统的语音识别部分,采用了线性预测(LPC)及矢量量化的距离(VQ)方法,过滤和识别人的语音信号;文本-语音转换部分,通过对每个动态打开的网页,采用文字转化成语音(text-to-speech,TTS)技术及语音合成平台完成语音合成,将其网页文本内容朗读出来,同时也可对用户进行语音提示。     

基于红外条码的盲人扑克游戏辅助仪的设计

设计了一款基于集条形码技术、嵌入式系统及语音芯片技术于一体、结构紧凑、操作简单方便的盲人扑克游戏辅助仪。该装置设计了常见的扑克游戏程序,使用者可根据兴趣方便地选择游戏种类。其主要特点是利用语音技术能够实时播报其他人的出牌信息并可根据需要耳机播报本人手中牌的信息,让盲人用"听"牌代替了"摸"牌,使盲人以及不懂盲文的低视力人群能像正常人一样打扑克、玩桥牌等,克服了现有低视力群体打牌游戏中常见的问题,如盲文识别错误等。     

基于边缘检测的盲人图像感知系统设计

采用图像的边缘检测技术,设计了盲人图像感知系统,使盲人通过皮肤电刺激来感知图像信息,为盲人提供了一种获取视觉信息的新方法。该系统包括图像采集与处理系统和电触觉装置,用于产生皮肤电刺激。盲人通过摄像机采集图像,对该图像进行边缘检测得到边缘图像。然后把边缘图像发送给电触觉装置的单片机微处理器,单片机根据接收到的边缘图像数据控制二维电极阵列产生相应的电刺激,盲人通过皮肤电触觉来感知图像的边缘轮廓信息。盲人图像感知系统使盲人能够以触觉的方式"看到"视觉图像,大大提高了盲人的生活质量。     

导盲系统的设计及实现

目前多数的视力患者通过使用盲人杖来辅助独立行走,但盲人杖的功能单一,且存在诸多安全隐患,针对以上问题文中提出的导盲系统在普通盲人手杖的基础上,利用红外测距原理,识别周围是否有物体,利用语音播报模块发出语音信号提示,利用温度传感器感应盲人手心位置温度变化辅助盲人找回导盲杖,通过GPS定位系统,有效地防止盲人走失,实现对盲...     

智能语音盲人拐杖设计方案

目前许多盲人出行极不方便,时常经历盲道被占用,路途有障碍,导盲犬不许上公交等情况。为了使他们的生活更加便利,文中设计了一款基于STC15F2K60S2单片机的智能语音盲人拐杖。本项目采用调查法,调查目标用户盲人群体的出行状况以及现存盲人智能拐杖的基本情况,了解盲人群体需求,综合调查资料,初步确定了软硬件设计思路。目前确定初步实现环境信息显示、语音识别、短信报警、GPS定位等功能。增加按键,控制不同的功能,实现功能的简单切换。按动按键,当在设定范围内出现障碍物时,系统语音助手自动播报预警语音;按下语音导航按键后,导航系统自动规划最优路线,语音助手播报位置并导航;当使用者遇到麻烦时,按下SOS按键,系统自动联系紧急联系人,紧急联系人若未接通电话,则自动转接距离最近的警局;当使用者准备上楼梯时,按下伸缩按键,拐杖自动伸缩,辅助使用者上下楼梯。该智能语音拐杖有助于提升盲人用户的生活质量,保障他们的出行安全。     

盲人手机软件面市

一款专门供盲人明友使用的手机软件近日在广东举办的残疾人用品用具展亮相。这款软件可以安装在具有上网等功能的高端手机中，通过软件的应用，可以将手机里菜单内容直接转化成语音，当盲人按下某个键时也会通过语音提示告诉盲人按下了什么。当盲人收到短信息时，这套语音软件也可以将内容“告诉”给盲人，他们甚至可以通过触摸按键发送短信息。当进入编辑短信功能时，盲人可以通过拼音等方式进行语音的组合，手机会用词组的方式告诉使用者选择的字词。当通过手机上网时，接收到的网络内容同样可以转换成语音输出，十分方便。     

文盲智能辅助系统的设计与实现

为了解决文盲弱势群体在生活中因识字困难所带来问题,本文设计了一款基于机器学习的文盲智能辅助系统。该系统使用Flask框架,结合并改进多个模型实现系统的各个功能。通过结合讯飞的语音听写流式接口和语音播报接口实现语音识别功能,利用paddle库的OCR模型结合讯飞的语音播报接口实现自然场景下的图片文字识别功能。经过实践表明,该系统在日常生活中能够为文盲弱势群体提供帮助。     

基于机器视觉的智能导盲眼镜设计

提出一种基于机器视觉的智能导盲眼镜系统的设计方案。采用三星公司Cortex-A8架构的S5PV210作为中央处理器,搭载Linux系统,配备双目采集、GPS定位、语音播报、GSM短信、语音通话、无线传输六大核心功能模块搭建智能导盲眼镜系统的硬件平台,结合深度学习算法在远程云服务器上完成了对目标场景的智能识别,最后以语音的形式实时对盲人的行走作出准确引导。系统测试结果表明,该智能导盲眼镜系统在测试环境下不仅能对盲人出行正确导航,还具有一定的目标识别能力,能帮助盲人进行简易物品归类。该系统还兼有GPS定位、语音通话、GSM短信等多项辅助功能。     

基于RFID的交互式盲杖设计

主要研究一种基于RFID的交互式盲杖,此交互式盲杖由两部分组成,盲杖部分由盲人持有,有源电子标签部分由盲人的亲朋好友所携带。适用人群是盲人及一些弱视群体。主要实现两大功能,避障和社交。利用传统的超声波装置实现避障,当前方2 m内有障碍物时通过语音提示盲人绕道行走;同时利用射频识别技术(RFID)识别周边4 m内的熟人,通过语音告诉盲人熟人的信息,以此实现盲人主动与熟人打招呼。相较于市面上广泛存在的智能盲人用品,此装置主要实现社交功能。     

Using ontology and RFID to develop a new Chinese Braille learning platform for blind students

Blind people must often rely on various types of Braille aids for reading. For blind students with little or no Braille-related knowledge, the use of an informational learning platform can effectively relieve the burden currently borne by their teachers or instructor. The OntoBraille@RFID (OBR) platform adopts ontological theory to retrieve Braille knowledge to facilitate learning, and a low-cost RFID for Braille character identification. This system is equipped with a voice device so students can learn independently. OBR is also equipped with administrative functions so parents can monitor the students’ learning progress. In contrast to currently available Braille learning devices, this new system can automatically expand the Braille database. Results from trials suggest that OBR can significantly improve the learning outcome for blind students. ORB has also obtained Chinese patent certification.     

基于Android平台的盲人手机系统设计与开发

人们生活中的绝大部分信息都是通过视觉获得的,所以盲人能够从外界获取的信息量很少,但这并不能阻止他们努力提高生活质量的渴望。智能手机的快速发展给盲人提供了前所未有的机遇。开发盲人手机具有极大的应用价值。该文主要介绍了在Android平台上设计和实现的一款盲人手机系统,该系统使用了语音识别、语音合成及Web Service等相关技术,实现了语音拨打电话、语音接听电话、语音发送短信、语音播报来电短信、语音报时、语音播报日期和语音播报天气等功能,并能在Android手机中稳定运行,方便盲人使用手机,具有较高的实用价值。     

Multilevel multisensory-based security system for an intelligent home

This article describes a multilevel multisensor-based security system that has multiple interfaces to be applied in an intelligent home. The security system contains four levels. There is a passive detection level, an active detection level, a system supervising level, and a remote supervising level. The control unit of these passive modules is a HOLTEK microchip. Each passive module has two different interfaces (wireless RF and voice). These modules can use a voice to alarm users than an event has occurred, and can transmit the real-time status and image signal to the active detection level and the system’s supervising level via the wireless RF interface. The active detection level can communicate with other levels via the wireless RF interface or the wireless Internet. The remote supervising level, the supervising level, and the active detection level can communicate with other levels via the wireless Internet. The status of these modules can also be displayed on the monitor of the supervising computer. Finally, we present some experimental results using passive and active detection modules in the security system.     

基于模糊控制的盲人避障系统设计

文章基于多个超声波传感器对地面信息的采集,利用模糊控制算法将采集的信息转化为指引盲人避障的语音提示。该系统采用超低功耗的MSP430单片机作为主控单元,外围设计了五路超声波传感器采集全方位的路面信息。通过分析路面信息,从而给出盲人前进方向以及前进的速度语音提示,实现盲人避障的功能。     

New Applications & Recent Research

The Kurzweil Reading Machine converts print to speech, and is designed as a reading tool for the blind and visually handicapped. The system handles ordinary printed material–books, letters, reports, memoranda, etc.–in most common styles and sizes of type. The output produced is a synthetic voice using full-word English speech. The reader operates the device by placing printed material face down on the glass plate forming the top surface of the scanning unit; he then presses the "page" button on the contol panel, and listens to the synthetic speech produced as an electronic camera scans the page and transmits its image to a minicomputer housed within the device. The computer separates the image into discrete character forms, recognizes the letters, groups the letters into words, computes the pronunciation of each word, and then produces the speech sounds associated with each phoneme. The machine operates at normal speech rates, about 150 words per minute.     

Building a pedagogy around action and emotion: experiences of Blind Opera of Kolkata

Contemporary knowledge systems have given too much importance to visual symbols, the written word for instance, as the repository of knowledge. The primacy of the written word and the representational world built around it is, however, under debate—especially from recent insights derived from cognitive science that seeks to bring back action, intent and emotion within the core of cognitive science (Freeman and Nunez in J Consciousness Stud 6(11/12), 1999). It is being argued that other sensory experiences, apart from the visual, along with desires (or intent) and emotions—like pain, pleasure, sorrow or joy—constitute equally important building blocks that shape an individual’s cognition of the world around. This multi-sensory cognition colored by emotions inspire action and hence is valid knowledge. This is probably nowhere more apparent than in the world of the visually impaired. Deprived of visual sensory capability, they have to perforce depend on other senses. But the dominant discourse in wider society plays a major role in determining what they (the blind) can do. A society built around visual symbols and the written word underplays other elements of cognition and in the process undervalues them. This also gets reflected in the construction of social artifacts of various kinds, such as the educational certification system (The Braille system is an attempt to make the written world accessible to the blind through tactile signals—so that words are ‘felt’ and ‘read.’ But it is quite cumbersome. For instance, even a blind highly skilled at writing in Braille would not be able to match the writing speed of an ordinary visually endowed literate person. Effective and efficient computer-based voice–text–voice converters might solve this problem better.)-based primarily on skills over the written word. Linguistic ability becomes most valuable and at another level the written word gets salience over the spoken word. The blind hardly has a chance, therefore, except through concessions or piety. A practice built around the imagery of an empowered blind person, therefore, must depart from mainstream conceptualization—for power is derived from what one has rather than from what one lacks. It must begin by tapping and valorizing one’s own endowments. This paper is an attempt to identify such a departure based on the experience of Blind Opera—a theatre group of the blind working in Kolkata, India. It seeks to provide an exposition in written word of an experience that can only be partially captured within the confines of a text. It is an incomplete account, therefore, and may be taken as an attempt to reach out and seek an exchange of experiences and insights.     

高档住宅区智能挡车器的设计

该智能挡车器基于RFID射频识别技术,并且具有图像抓拍和语音提示功能.提出了系统完整设计方案,并对系统硬件电路和软件模块进行介绍.     

Effect of outside view on attentiveness in using see-through type augmented reality device

When a viewer watches a virtual image through a see-through augmented reality (AR) device and an outside view simultaneously, the viewer’s attentiveness is affected. In this study, the authors examined the relationship between the viewer’s attentiveness and viewing distance of the outside view. The virtual image in the see-through AR device was at a fixed distance of 2.5 m, while a TV at distances of 0.33, 1, 2.5 and 5 m was considered as the outside view. In this experiment, thirty participants viewed a movie on an AR device and ten target images of pictures and words were sequentially displayed on the TV by overlapping them onto the virtual image. The participants were then asked to identify these target images from examples of sixty images. A correct answer (CA) ratio was used to measure the participant’s attentiveness. Results demonstrated that CA ratios at distances of 0.33 and 1 m was better than CA ratios at distances of 2.5 and 5 m. Further, the CA ratio of target images of pictures was better than that of words. When the virtual and target images were located at similar distances, these images were sharply focused on the retina. However, when the distances of these images were different, the participant’s eyes focused on one image causing a defocus on the other. A better CA ratio at different distances implied that the difference of sharpness in overlapped virtual and target images helped the participants in identifying the virtual or target image rather than two overlapped images of similar sharpness. Thus, sharply focused virtual and target images at the same distance did not guarantee better identification of two distinct images.     

基于DCT域实时语音的信息隐藏方法

为了满足对秘密信息的实时性的需求,提出一种基于DCT域实时语音的信息隐藏方法.该方法使用DirectSound 对实时的语音进行捕获,依据人类听觉系统的掩蔽效应,选择DCT中的高频系数,通过量化秘密的音频信息将其嵌入到捕获的原始语音中,通过Socket技术对语音进行实时的传输,并且在秘密信息的提取过程中不需要原始音频载体信息的参与,实现秘密信息的盲提取.实验结果表明,在网络较好的情况下可以有效地借助实时的语音传递秘密信息.