Method for verification of signatures and handwriting based on comparison of extracted features

• US 5,995,953 A
• Filed: 07/16/1997
• Issued: 11/30/1999
• Est. Priority Date: 02/19/1993
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Rindtorff, Klaus, Rudolph, Volker
• Primary Examiner(s): Downs, Robert W.
• Application Number:
  US08/895,335
• Time in Patent Office: 867 Days
• Field of Search: 395/23, 395/24, 382/156, 382/187, 382/186
• US Class Current: 706/20
• CPC Class Codes: G06V 40/30 Writer recognition; Reading and verifying signatures
• CPC Class Codes: G07C 9/35 by means of a handwritten signature

Abstract

This invention describes a technology to improve the feature based comparison of images. The images are captured and their significant features are extracted. For a comparison only the feature values have to be compared instead of the images themselves. This leads to a significant reduction of storage space and calculation time needed. The reliability of the comparison is improved greatly by including also the individual variation ranges of the feature values and using a specialized neural net for classification.

Claims

1. Multi-layer neural net computer for comparing features from a sample image and a stored master reference, comprising:

an image input to said computer for capturing said sample image;

a store communicating with said computer for storing said master reference; and

a multi-layer neural net further comprising;

an input layer of said net wherein input to each node in said layer includes a feature value from said sample image to be verified, a feature value from said stored master reference, and a feature range value from said stored master reference; and

wherein output from each node in said layer includes a distance value for the sample image feature value wherein said distance value is calculated from said input to each node.

2. Multi-layer neural net computer as in claim 1, wherein said distance value is computed by a function having its extremum for two equal feature values.

3. Multi-layer neural net computer as in claim 2, wherein said function is computed according to the formula ##EQU3## wherein abs is the absolute value of S-R,

S is the feature value from the sample image,

R is the feature value from the stored master reference, and

V is the feature range value from the stored master reference.

4. Multi-layer neural net computer as in claim 1, wherein features are contained in two classes representing values from geometric and stylistic information in the image.