Neural network system for classifying fingerprints

• US 5,825,907 A
• Filed: 07/11/1997
• Issued: 10/20/1998
• Est. Priority Date: 12/28/1994
• Status: Expired due to Term

• Current Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Russo, Anthony Peter
• Primary Examiner(s): Couso, Jose L.
• Assistant Examiner(s): Do, Anh Hong
• Application Number:
  US08/893,668
• Time in Patent Office: 466 Days
• Field of Search: 382/124, 382/125, 382/156, 382/181, 382/170, 382/224, 395/24, 395/21
• US Class Current: 382/124
• CPC Class Codes: G06V 40/1359 Extracting features related to ridge properties; Determining the fingerprint type, e.g. whorl or loop

Abstract

A system for automatically classification of human fingerprints. An unidentified fingerprint is processed to produce a direction map. The direction map is processed to generate a course direction map. The coarse direction map is input to a locally connected, highly constrained feed-forward neural network. The neural network has a highly structured architecture well-suited to exploit the rotational symmetries and asymmetries of human fingerprints. The neural network classifies the unidentified fingerprint into one of five classifications: Whorl, Double Loop, Left Loop, Right Arch and Arch.

Claims

1. A system for classifying an unknown fingerprint image having a plurality of pixels, comprising,

means for generating a pixel direction map from ridge directions found in the unknown fingerprint image;

means for generating a coarse direction map by dividing said pixel direction map into regions and determining for each of the regions a coarse direction as a most frequent pixel direction within the region; and

a neural network classifier having an input layer of neurons connected to receive said coarse direction map and connected to a hidden layer of neurons, wherein square regions of said input layer of neurons are arranged in a plurality or rings and connected to said hidden layer of neurons, said neural network classifier further includes an output layer of neurons that represent a plurality of fingerprint classifications.

2. The system of claim 1, wherein said neural network classifier classifies the unknown fingerprint image into either a left loop, a right loop, an arch, a whorl or a double whorl.

3. The system of claim 1, wherein said neural network classifier is a locally connected, highly constrained feed-forward neural network.

4. The system of claim 1, wherein said square regions within each of said plurality of rings overlap.

5. The system of claim 1, wherein global sections of said input layer of neurons are connected to said hidden layers of neurons, wherein said global sections are generated by dividing said coarse direction map into four equal quadrants.

6. The system of claim 1, wherein said input layer of neurons is connected to said hidden layer of neurons via a plurality of weights, each of said square regions within each of said plurality of rings from said input layer of neurons is connected via the same set of said plurality of weights to said hidden layer of neurons.

7. The system of claim 1, wherein said neural network classifier has at least two hidden layers.

8. The system of claim 7, wherein said at least two hidden layers include a first hidden layer connected to a second hidden layer using weight sharing.

9. The system of claim 8, wherein said neural network classifier evaluates the unknown fingerprint image from four different distances from the center of the unknown fingerprint image.

10. A system for classifying an unknown image having radial characteristics and a plurality of pixels, comprising,

means for generating a pixel direction map from ridge directions found in the unknown image;

means for generating a coarse direction map by dividing said pixel direction map into regions and determining for each of the regions a coarse direction as a most frequent pixel direction within the region; and

a neural network classifier having an input layer of neurons connected to receive said coarse direction map and connected to a hidden layer of neurons, wherein square regions of said input layer of neurons are arranged in a plurality or rings and connected to said hidden layer of neurons, said neural network classifier further includes an output layer of neurons that represent a plurality of fingerprint classifications.