### Abstract

The authors consider a prototype-based character recognizer that makes comparisons based on blurred representations of the images. The blurring induces a metric on the space of all images that varies continuously under continuous deformations of the image plane. This blurred representation is suitable for direct implementation of a nearest neighbor classifer. However, it is still desirable to have a representation which is invariant under rotation, translation, and scaling of the image plane. A representation which is locally invariant under these transformations is produced by transforming an input to a local minimum of its distance from each prototype simultaneously. These minima are found by performing a gradient descent on an appropriate error surface over the four transformation parameters. The error functional is the L_{2}-norm of the difference between the blurred prototype and the blurred input. The resulting classifier makes more efficient use of prototypes than the nearest neighbor classifier.

Original language | English (US) |
---|---|

Title of host publication | Proc 1988 IEEE Int Conf Syst Man Cybern |

Editors | Anon |

Pages | 361-364 |

Number of pages | 4 |

State | Published - 1988 |

Event | Proceedings of the 1988 IEEE International Conference on Systems, Man, and Cybernetics - Beijing/Shenyang, China Duration: Aug 8 1988 → Aug 12 1988 |

### Other

Other | Proceedings of the 1988 IEEE International Conference on Systems, Man, and Cybernetics |
---|---|

City | Beijing/Shenyang, China |

Period | 8/8/88 → 8/12/88 |

### Fingerprint

### ASJC Scopus subject areas

- Engineering(all)

### Cite this

*Proc 1988 IEEE Int Conf Syst Man Cybern*(pp. 361-364)

**Handwritten character recognition using a gradient classifier.** / Brandt, Robert D.; Wang, Yao; Laub, Alan J.; Mitra, Sanjit K.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

*Proc 1988 IEEE Int Conf Syst Man Cybern.*pp. 361-364, Proceedings of the 1988 IEEE International Conference on Systems, Man, and Cybernetics, Beijing/Shenyang, China, 8/8/88.

}

TY - GEN

T1 - Handwritten character recognition using a gradient classifier

AU - Brandt, Robert D.

AU - Wang, Yao

AU - Laub, Alan J.

AU - Mitra, Sanjit K.

PY - 1988

Y1 - 1988

N2 - The authors consider a prototype-based character recognizer that makes comparisons based on blurred representations of the images. The blurring induces a metric on the space of all images that varies continuously under continuous deformations of the image plane. This blurred representation is suitable for direct implementation of a nearest neighbor classifer. However, it is still desirable to have a representation which is invariant under rotation, translation, and scaling of the image plane. A representation which is locally invariant under these transformations is produced by transforming an input to a local minimum of its distance from each prototype simultaneously. These minima are found by performing a gradient descent on an appropriate error surface over the four transformation parameters. The error functional is the L2-norm of the difference between the blurred prototype and the blurred input. The resulting classifier makes more efficient use of prototypes than the nearest neighbor classifier.

AB - The authors consider a prototype-based character recognizer that makes comparisons based on blurred representations of the images. The blurring induces a metric on the space of all images that varies continuously under continuous deformations of the image plane. This blurred representation is suitable for direct implementation of a nearest neighbor classifer. However, it is still desirable to have a representation which is invariant under rotation, translation, and scaling of the image plane. A representation which is locally invariant under these transformations is produced by transforming an input to a local minimum of its distance from each prototype simultaneously. These minima are found by performing a gradient descent on an appropriate error surface over the four transformation parameters. The error functional is the L2-norm of the difference between the blurred prototype and the blurred input. The resulting classifier makes more efficient use of prototypes than the nearest neighbor classifier.

UR - http://www.scopus.com/inward/record.url?scp=0024169104&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0024169104&partnerID=8YFLogxK

M3 - Conference contribution

SN - 7800030393

SP - 361

EP - 364

BT - Proc 1988 IEEE Int Conf Syst Man Cybern

A2 - Anon, null

ER -