Misc.

May 19, 2000

Consideration for bar code position detection using a neural network with mutual inhibitions : Application to work assistance for visually impaired

IEICE technical report. ME and bio cybernetics
  • HAYASAKA Shuichi
  • ,
  • MAEDA Yoshinobu
  • ,
  • MAKINO Hideo
  • ,
  • YAMAMOTO Satoshi
  • ,
  • SEKIYA Yasushi

Volume
100
Number
98
First page
109
Last page
116
Language
Japanese
Publishing type
Publisher
The Institute of Electronics, Information and Communication Engineers

We have developed an assistive system for the visually impaired to recognize functional parts of objects. The necessary information for the visually impaired is encoded in a two-dimensional (2D) code that put on the object. In our study the 2D code is "invisible" in the sense that it is hidden in the background design, though it is visible in the near infrared spectrum. Therefore, we cannot recognize it through our naked eyes. We extended the faculty of our system in this report so that the system could recognize successively the several 2D codes. To this end, we used an artificial neural network in which each neuron model is connected by means of mutual inhibitions. A static image (640×480 pixels) obtained by the CCD camera was given to the neural network composed of 357 Nagumo-Sato neuron models. The neural network was set so as to overspread the image. When the network parameters were regulated at appropriate values, the corresponding neurons to the location of 2D codes could generate randomly a bursting excitation their turns. A 2D code is 1.5×1.5cm in area. The distance from CCD camera to 2D code is 40 cm. When initial values of the neuron models were identical, we confirmed that our system could detect five 2D codes during forty seconds by using numerical simulation (Pentium II prcessor, 450MHz).

Link information
CiNii Articles
http://ci.nii.ac.jp/naid/110003288090
CiNii Books
http://ci.nii.ac.jp/ncid/AN1001320X
URL
http://id.ndl.go.jp/bib/5410233
URL
http://search.jamas.or.jp/link/ui/2000267976
ID information
  • ISSN : 0913-5685
  • CiNii Articles ID : 110003288090
  • CiNii Books ID : AN1001320X

Export
BibTeX RIS