Presentation | 1999/3/18 Noise Immunity Characteristics for Object Position and Shape Recognition by Spreading Associative Neural Network Hirokazu KANAYAMA, Kiyomi NAKAMURA, Tadataka YOSHIKAWA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Paying attention to the parietal cortex, a spreading associative neural network was proposed that can simultanously recognize the object position and shape. The spreading associative neural network net only recognizes the position of an object irrespective of its shape, but also recognizes its shape irrespective of its position, in both the learned and unlearned positions. In this research, the noise immunity characteristic of the spreading associative neural network is studied as to object position and shape rccognition. The objects (images) are binary Arabic numerals (20×20dot) or gray scaled human faces (80×80dot). Various random dot noises are added in the recollection step and the recognition characteristics are examined. As to the shape recognition, an Arabic numeral can be recognized till 15%-20% noise rate, and a gray scaled human face can be recognized till 35%-40% noise rate. The average positional recognition error is 4.6 dots for the Arabic numeral at 55% noise rate, and it is 15.6 dots for the gray scaled human face at 85% noise rate. The noise immunity characteristic is better for the human face than far the binary Arabic numerals, furthermore, it is better for the positional recognition than for the shape recognition. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Noise immunity characteristic / Spatial spreading / Gray scaled human face / Binary Arabic numeral |
Paper # | NC98-141 |
Date of Issue |
Conference Information | |
Committee | NC |
---|---|
Conference Date | 1999/3/18(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Neurocomputing (NC) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Noise Immunity Characteristics for Object Position and Shape Recognition by Spreading Associative Neural Network |
Sub Title (in English) | |
Keyword(1) | Noise immunity characteristic |
Keyword(2) | Spatial spreading |
Keyword(3) | Gray scaled human face |
Keyword(4) | Binary Arabic numeral |
1st Author's Name | Hirokazu KANAYAMA |
1st Author's Affiliation | Department of Electronics and Informatics, Toyama Prefectural University() |
2nd Author's Name | Kiyomi NAKAMURA |
2nd Author's Affiliation | Department of Electronics and Informatics, Toyama Prefectural University |
3rd Author's Name | Tadataka YOSHIKAWA |
3rd Author's Affiliation | Department of Electronics and Informatics, Toyama Prefectural University |
Date | 1999/3/18 |
Paper # | NC98-141 |
Volume (vol) | vol.98 |
Number (no) | 673 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |