On-line learning for neural net-based character recognition systems

• US 5,966,460 A
• Filed: 03/03/1997
• Issued: 10/12/1999
• Est. Priority Date: 03/03/1997
• Status: Expired due to Term

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Porter, Gilbert B. III, Roberts, Frederick J. Jr., Fan, Zhigang
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): PATEL, KANJIBHAI B
• Application Number:
  US08/810,388
• Time in Patent Office: 953 Days
• Field of Search: 382/161, 382/155, 382/156, 382/157, 382/158, 382/159, 382/165, 382/190, 382/181, 382/187, 382/224, 382/225, 382/226, 382/227, 382/228
• US Class Current: 382/159
• CPC Class Codes: G06F 18/24137 Distances to cluster centroïds
• CPC Class Codes: G06V 30/10 Character recognition

Abstract

A neural network based improving the performance of an omni-font classifier by using recognized characters for additional training is presented. The invention applies the outputs of the hidden layer nodes of the neural net as the feature vector. Characters that are recognized with high confidence are used to dynamically train a secondary classifier. After the secondary classifier is trained, it is combined with the original main classifier. The invention can re-adjust the partition or boundary of feature space, based on on-line learning, by utilizing the secondary classifier data to form an alternative partition location. The new partition can be referred to when a character conflict exists during character recognition.

Claims

1. A character recognition system, comprising:

means for capturing characters to be analyzed;

an artificial neural network system including at least one primary classifier comprising a primary classification memory means and at least one secondary classifier comprising a secondary classification memory means, said artificial neural network system learning a first set of characters and storing the first set of characters in said primary classifier and said artificial neural network system storing a second set of characters related to the first set of characters in said secondary classifier; and

a character analyzer for analyzing candidate characters captured by said capturing means, said character analyzer causing candidate characters to be compared with said first set of characters stored in said primary classifier when said candidate characters are recognizable with a high degree of confidence said high degree of confidence being used to dynamically train said secondary classifier, and said character analyzer causing candidate characters to be compared with said second set of characters stored in said secondary classifier when said candidate characters are recognizable with a low degree of confidence.

2. The character recognition system of claim 1, in which a candidate character matches a character in one of the first set and the second set, further comprising a character rendering subsystem for rendering the candidate character which matches the character in the one of the first set and the second set.

3. The character recognition system of claim 1, wherein the second set of characters in the second classification memory are learned twice in said artificial neural network.