BIOMETRIC TRAINING AND MATCHING ENGINE

• US 20130083975A1
• Filed: 10/03/2012
• Published: 04/04/2013
• Est. Priority Date: 10/03/2011
• Status: Active Grant

• Current Assignee: Accenture Global Services Limited (Accenture PLC)
• Original Assignee: Accenture Global Services Limited (Accenture PLC)
• Inventors: PARTINGTON, Alastair Ross, TORNESELLO, Sebastiano, ROCAMORA MARTI, Jordi
• Granted Patent: US 9,053,374 B2
• Time in Patent Office: Days
• US Class Current: 382/115
• CPC Class Codes: G06F 16/90335 Query processing
• CPC Class Codes: G06V 10/94 Hardware or software architectures specially adapted for image or video understanding
• CPC Class Codes: G06V 40/50 Maintenance of biometric data or enrolment thereof
• CPC Class Codes: G06V 40/70 Multimodal biometrics, e.g. combining information from different biometric modalities

Abstract

The present disclosure concerns a method of identifying a biometric record of an individual in a database (108) having a plurality of biometric records, the method involving: during a training phase: applying by a processing device a matching operation to determine scores for a similarity between at least one training biometric sample of each of a plurality of training records and at least one probe sample; based on said scores, determining a threshold value (S_TH,M_TH); and during an identification phase: evaluating at least one reference biometric sample of each of the records of said database to determine a parameter value for each record; selecting a subset of said records by comparing each of said parameter values with said threshold value; and applying a matching operation to the selected records to determine whether an input biometric sample matches a reference biometric sample of one of said selected records.

Claims

1. A method of identifying a biometric record of an individual in a database (108), the database comprising a plurality of biometric records, the method comprising:

during a training phase;

applying by a processing device (302) a matching operation to determine scores for a similarity between at least one training biometric sample of each of a plurality of training records and at least one probe sample;

based on said scores, determining by said processing device a threshold value (S_TH,M_TH); and

during an identification phase;

evaluating at least one reference biometric sample of each of the records of said database to determine a parameter value for each record;

selecting a subset of said records by comparing each of said parameter values with said threshold value; and

applying a matching operation to the selected records to determine whether an input biometric sample matches a reference biometric sample of one of said selected records.

2. The method of claim 1, wherein said threshold value is determined by said processing device by evaluating a correlation between said scores and at least one parameter value of each record.

3. The method of claim 1, wherein a plurality of parameter values are evaluated for each record, and said training phase further comprises selecting which of the plurality of parameter values is to be compared with said threshold value.

4. The method of claim 1, wherein said training phase comprises evaluating said at least one training biometric sample of each training record to determine a value of said parameter for each training record, said threshold value being determined as a parameter value that allows known matching and non-matching samples to be separated.

5. The method of claim 4, wherein said parameter provides a quality indication of said sample.

6. The method of claim 5, wherein said parameter is based on either:

the age of the individual of the biometric sample;

or

the gender of the individual of the biometric sample;

or

the sharpness of an image of the biometric sample;

or

the viewing angle of an image of the biometric sample;

or

the contrast in an image of the biometric sample;

or any combination thereof.

7. The method of claim 1, wherein selecting a subset of said records comprises creating a first partition of said database (108) comprising said selected records, the method further comprising creating at least one further partition of said database comprising the remaining records of said database.

8. The method of claim 7, wherein said identifying phase further comprises applying a further matching operation to the records of the at least one further partition of said database to determine whether said input biometric sample matches a reference biometric sample of one of said records of the further partition.

9. The method of claim 1, wherein said training records each comprise a plurality of training samples of different types from each other, and wherein said training phase further comprises determining, based on said scores, an order that said matching operation is applied to said sample types.

10. The method of claim 1, wherein said training phase further comprises pseudo-randomly selecting the plurality of training records from among the plurality of biometric records of said database.

11. A computer readable medium storing a computer program that, when executed by a processor, causes the method of claim 1 to be implemented.

12. A biometric matching device comprising:

a database (108) storing a plurality of biometric records;

a memory (114) storing a plurality of training records, each comprising at least one training biometric sample, and at least one probe sample; and

a processor (302) configured to;

apply during a training phase a matching operation to determine scores for a similarity between at least one training biometric sample of each of a plurality of training records and at least one probe sample, and based on said scores, determining by said processing device a threshold value (S_TH ,M_TH); and

evaluate during an identification phase at least one reference biometric sample of each of the records of said database to determine a parameter value for each record, select a subset of said records by comparing each of said parameter values with said threshold value; and

apply a matching operation to the selected records to determine whether an input biometric sample matches a reference biometric sample of one of said selected records.

13. The biometric matching device of claim 12, wherein said processor is configured to determine said threshold value by evaluating a correlation between said scores and at least one parameter value of each record.

14. The biometric matching device of claim 12, wherein said processor is further configured to create a first partition of said database comprising said selected records, and to create at least one further partition of said database comprising the remaining records of said database.

15. A biometric matching system comprising:

one or more biometric capturing devices (104,106) for capturing at least one biometric sample of an individual; and

the biometric matching device of claim 12.