【作者】 向彪；

【导师】 戴士杰；

【作者基本信息】 河北工业大学 ， 机械工程， 2011， 硕士

【摘要】 随着世界盲人数量的不断增加和盲人群体受关注程度的不断提高，各种助盲设备也应运而生。而电子行走辅助系统凭借其携带方便，结构简单，易使用的优势，正日益成为助盲技术研究的主流。传统的助盲系统传达给盲人的信息只有障碍物的方位和位置信息，而无法将障碍物的具体特征告知盲人形成视觉重现。针对这一缺陷，本文构建了基于超声波和视觉信息融合的语音提示系统，充分利用盲人的先验知识，很好的实现了外界环境障碍物信息的实时再现。在图像识别方面，为了克服在外界环境变化时单一图像特征有可能失效的缺陷，采用超声波和视觉信息相融合的方法来进行图像识别。运用超声波阵列探测，根据最小距离判定法则，判断出最近障碍物的方位。从而驱动摄像头旋转一定角度拍摄图像，滤波后，提取出图像的颜色和形状特征。根据不同物体的颜色、形状特征对于不同物体的敏感程度不同，先用支持向量机对提取出来的单一特征进行分类识别，得出特征权重计算因子，从而定义特征权重计算方法。引入到K近邻分类器距离函数中，将颜色、形状特征融合起来，结合图像数据库，能够很好的实现物体识别。在语音提示方面，事先将要播放的语音内容分段录制在语音芯片中，并由LCD显示单元读出分段内容所处的地址。上位机将图像识别的结果传输给语音提示电路中的单片机，通过触发不同的语音地址，从而实现语音的连续播放。单片机中的程序在KEIL C软件中编写并调试，并通过运行STC_ISP软件将程序下载到单片机中。完成下载后，单片机自动运行程序。在上述研究基础上搭建语音提示系统平台，并就图像的声音提示效果进行了测试和评估。实验证明，在盲人有先验知识的前提下，该语音提示系统能很好的帮助盲人实现视觉再现。更多还原

【Abstract】 With the increasing number of the world’s blind and more attention to them, a variety ofdevices came into being to help the blind. Electronic travel aid system, with its easy to carry,simple structure, easy to use advantages, is increasingly becoming the mainstream technology tohelp the blind.The traditional system to help the blind to convey to the blind only obstacle locationinformation and location information, can not inform the blind of specific characteristics of theobstacle to form visual reproduction. For this defect, we construct a voice prompts system basedon ultrasonic and visual information fusion. It fully utilizes the prior knowledge of the blind, andrealized the external environment obstacle information real-time reappearance.In image recognition, in order to overcome the defect that single feature may fail when theexternal environment change, we use the approach of the fusion of ultrasonic and visualinformation to image recognition. Using the ultrasonic wave array survey, according to theminimum range determination principle, judges the orientation of the recent obstacles. That drivethe rotation angle shoted images, filtered, withdrawed the image the color and the shapecharacteristic. According to the object color, shape features in different objects have differentsensitivities, first uses support vector machines to the sole characteristic which withdraws carrieson the classified recognition, and obtains the characteristic weight computation factor, therebydefining feature weighting methods. K nearest neighbor classifier is introduced into the distancefunction, integrated the color and shape features, combined with image database, can achieve theobject recognition.In the voice prompt aspect, the pronunciation content partition record which is going tobroadcast beforehand on the voice chip by segment LCD display unit to read out the contents ofwhich the address. PC will transfer the image recognition results to the voice promptsmicrocontroller circuit, and realize a continuous playback of voice by triggering different voiceaddresses. The procedure was compiled and debuged in KEIL C software, and was downloadedto the microcontroller by running STC-ISP software. Completed, the microcontroller automatically run the program.Based on these studies to bulid the voice prompt system platform, and the voice prompts forthe image effects were tested and evaluated. Experiments show that a prior knowledge in theblind context, the voice prompts system to help blind people achieve good visual representation.更多还原