Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD FOR JUDGING PROCESSOR BASE
Document Type and Number:
Japanese Patent JP3272842
Kind Code:
B2
Abstract:

PURPOSE: To recognize a character string without requiring the detection or recognition of characters by generating one set of features based on a word form in a text line boundary box, applying one set of these features to a prescribed network and determining whether any specified text line includes the keyword to be searched or not.
CONSTITUTION: An input image is operated by a word boxer 30 for discriminating word boundary boxes related to all the latent keywords of the image. Concerning the image of latent keywords in the respective boundary boxes, word structure is analyzed for classifying the latent keywords and operated by a feature generator 32 for generating feature vectors at the horizontal positions at equal intervals. The set of these latent keywords is applied to a word image spotter 35. This word image spotter 35 is a network for receiving a word model. This word model is generated by linking sets 45 of character models.


Inventors:
Francine Earl Chen
Lindy Willcox
Dan S. Bloomberg
Application Number:
JP30926593A
Publication Date:
April 08, 2002
Filing Date:
December 09, 1993
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
Xerox Corporation
International Classes:
G06F17/30; G06T7/00; G06V30/262; G06V30/10; (IPC1-7): G06K9/70; G06F17/30; G06K9/50; G06K9/72; G06T7/00
Domestic Patent References:
JP4329597A
JP394299A
JP1265378A
JP684008A
JP3225488A
Other References:
中川聖一 大黒慶久 橋本泰秀,構文解析駆動型日本語連続音声認識システム,電子情報通信学会論文誌 D−2,日本,電子情報通信学会,1989年8月25日,Vol.J72−D−2 No.8,p.1276−1283
川端豪 花沢利行 鹿野清宏,HMM音韻認識に基づくワードスポッティング,電子情報通信学会技術研究報告 SP88−21〜30,日本,電子情報通信学会,1988年6月24日,SP88−23 Vol.88 No.92,p.17−22
甲斐充彦 中川聖一,ワードスポッティング方を用いたHMM連像音声認識についての検討,電子情報通信学会技術研究報告 SP92−7〜12,日本,電子情報通信学会,1992年5月22日,SP92−10 Vol.92 No.35,p.25−32
Attorney, Agent or Firm:
Atsushi Nakajima (1 person outside)