System and method for compressing biometric models

• US 6,580,814 B1
• Filed: 07/31/1998
• Issued: 06/17/2003
• Est. Priority Date: 07/31/1998
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ittycheriah, Abraham P., Maes, Stephane H.
• Primary Examiner(s): AHMED, SAMIR ANWAR
• Application Number:
  US09/126,894
• Time in Patent Office: 1,782 Days
• Field of Search: 382/115, 382/119, 382/120, 704/221, 704/222, 704/230, 704/231, 704/236, 704/237, 704/238, 704/239, 704/243, 704/245, 704/247, 704/250, 704/500, 704/246, 704/273, 235/380, 340/5--, 340/5--, 738/654
• US Class Current: 382/115
• CPC Class Codes: G06F 18/28 Determining representative reference patterns, e.g. by averaging or distorting; Generating dictionaries
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
• CPC Class Codes: G10L 17/04 Training, enrolment or model building

Abstract

A system and method for building compressed biometric models and performing biometric identification using such models. The use of the compressed biometric models results in a significant decrease in the storage requirements for biometric models in conventional biometric systems. A given number of L reference biometric models are built. The L reference models are randomly divided into M subsets. During user enrollment, distance measurements between a temporary biometric model and each of the reference models in the M subsets are computed. The rank and distance parameters are used to build the compressed biometric models in accordance with the model: I(M_j, R_i, D(M_j, R_i)), where I represents the identity of the closest reference model in a corresponding subset M_j; R_i refers to the ranking of the closeness of the reference model to the temporary biometric model as compared with the closeness of each of the other reference models in the corresponding subset M_j; and D refers to the corresponding distance measure between the reference model and the temporary biometric model.

Claims

1. A system for compressing biometric models, comprising:

means for receiving and processing biometric data from a plurality of reference individuals to generate a reference biometric model for each of said plurality of reference individuals and from at least one user of said system to generate a temporary biometric model for said user;

means for partitioning said plurality of reference biometric models into M subsets;

means for computing a distance between said temporary biometric model and each of said reference biometric models in said M subsets and, for each of said M subsets, identifying the reference biometric model which is closest to said temporary biometric model based on said computed distances; and

means for constructing a compressed biometric model for said user, wherein the compressed biometric model comprises a pattern of parameters, wherein the pattern of parameters comprises identity parameters corresponding to said closest reference models from said M subsets.

2. The system of claim 1, wherein said pattern of parameters comprises a pattern [I(M₁), . . . I(M_j)], where I is said identity of the closest reference model in the corresponding M_j^th subset.

3. The system of claim 1, wherein said pattern of parameters further comprises distance parameters corresponding to said computed distance between said identified closest reference models and said temporary biometric model.

4. The system of claim 3, wherein said pattern of parameters comprises a pattern {I(M₁, D(M₁)), . . . I(M_j, D(M_j))} where I represents said identity of the closest reference model in the corresponding M_j^th subset of said M subsets and D(M_j) represents said computed distance between said closest reference model in the corresponding M_j^th subset and said temporary biometric model.

5. The system of claim 1, wherein said computing means identifies and ranks R closest reference models in each corresponding one of said M subsets, and wherein said pattern of parameters further comprises ranking parameters corresponding to said ranking of each of said R closest reference models.

6. The system of claim 5, wherein said pattern of parameters comprises a pattern {[I(M₁, R₁), . . . I(M₁, R_i)], . . . , [I(M_j, R₁), . . . I(M_j, R_i)]}, where I is said identity of the R_i^th closest reference model in the corresponding M_j^th subset of said plurality of M subsets.

7. The system of claim 5, wherein said pattern of parameters further comprises distance parameters corresponding to said computed distance between each of said R closest models and said temporary biometric model.

8. The system of claim 7, wherein said pattern of parameters comprises a pattern {[I(M₁, R₁, D(M₁, R₁)), . . . I(M₁, R_i, D(M₁, R_i))], . . . , [I(M_j, R₁, D(M_j, R₁)), . . . , I(M_j, R_i, D(M_j, R_i))]}, where I represents said identity of the R_i^th closest reference model in the corresponding M_j^th subset of said plurality of M subsets and D represents said corresponding computed distance.

9. The system of claim 1, wherein said system is a text-independent speaker recognition system.

10. The system of claim 1, wherein said compressed biometric model is used as one of an encryption key, a decryption key, and a combination thereof.

11. The system of claim 1, further comprising:

means for storing a plurality of said compressed biometric models; and

means for comparing said stored compressed biometric models with a temporary compressed biometric model of a person during an identification and verification phase of said system, said temporary compressed biometric model being constructed by said constructing means, whereby said person will be identified if said comparison means finds a match between said temporary compressed biometric model and one of said stored compressed biometric models.

18. The system of claim 5, wherein R is a programmable parameter.

12. A method for building compressed biometric models in a biometric system, comprising the steps:

generating a plurality of reference biometric models;

partitioning said plurality of reference biometric models into M subsets;

generating a temporary biometric model for a user of said biometric system;

computing a distance between said temporary biometric model and each of said reference biometric models in said M subsets;

identifying the reference biometric model in each of said M subsets which is closest to said temporary biometric model based on said computed distances; and

constructing a compressed biometric model for said user, wherein the compressed biometric model comprises a pattern of parameters, wherein the pattern of parameters comprises identity parameters corresponding to said closest reference models from said M subsets.

13. The method of claim 12, wherein said pattern of parameters comprises a pattern [I(M₁), . . . I(M_j)], where I is said identity of the closest reference model in the corresponding M_j^th subset.

14. The method of claim 12, wherein said pattern of parameters further comprises distance parameters corresponding to said computed distance between said identified closest reference models and said temporary biometric model.

15. The method of claim 14, wherein said pattern of parameters comprises a pattern {I(M₁, D(M₁)), . . . I(M_j, D(M_j))} where I represents said identity of the closest reference model in the corresponding M_j^th subset of said M subsets and D(M_j) represents said computed distance between said closest reference model in the corresponding M_j^th subset and said temporary biometric model.

16. The method of claim 12, further comprising the steps of:

ranking the reference models in each corresponding one of said M subsets in accordance with the closeness of said reference models to the temporary biometric model, wherein said pattern of parameters further comprises ranking parameters corresponding to said ranking of said reference models in each of said M subsets.

17. The method of claim 16, wherein said pattern of parameters comprises a pattern {[I(M₁, R₁), . . . I(M₁, R_i)], . . . , [I(M_j, R₁), . . . I(M_j, R_i)]}, where I is said identity of an R_i^th closest reference model in the corresponding M_j^th subset of said plurality of M subsets.

19. The method of claim 16, wherein said pattern of parameters further comprises distance parameters corresponding to said computed distance between said identified and ranked reference models and said temporary biometric model.

20. The method of claim 19, wherein said pattern of parameters comprises a pattern {[I(M₁, R₁, D(M₁, R₁)), . . . I(M₁, R_i, D(M₁, R_i))], . . . , [I(M_j, R₁, D(M_j, R₁)), . . . , I(M_j, R_i, D(M_j, R_i))]}, where I represents said identity of the R_i^th closest reference model in the corresponding M_j^th subset of said plurality of M subsets and D represents said corresponding computed distance.

21. The method of claim 12, further comprising the steps of:

storing a plurality of said compressed biometric models; and

generating a temporary compressed biometric model of an individual making an identity claim;

comparing said stored compressed biometric models with said temporary compressed biometric model; and

verifying the identity of said individual if a match if found between said temporary compressed biometric model and one of said stored compressed biometric models.

22. A program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for building compressed biometric models in a biometric system, the method steps comprising:

generating a plurality of reference biometric models;

partitioning said plurality of reference biometric models into M subsets;

generating a temporary biometric model for a user of said biometric system;

computing a distance between said temporary biometric model and each of said reference biometric models in said M subsets;

identifying the reference biometric model in each of said M subsets which is closest to said temporary biometric model based on said computed distances; and

constructing a compressed biometric model for said user, wherein the compressed biometric model comprises a pattern of parameters, wherein the pattern of parameters comprises identity parameters corresponding to said closest reference models from said M subsets.

23. The program storage device of claim 22, wherein said pattern of parameters comprises a pattern [I(M₁), . . . I(M_j)], where I is said identity of the closest reference model in the corresponding M_j^th subset.

24. The program storage device of claim 22, wherein said pattern of parameters further comprises distance parameters corresponding to said computed distance between said identified closest reference models and said temporary biometric model.

25. The program storage device of claim 24, wherein said pattern of parameters comprises a pattern {I(M₁, D(M₁)), . . . I(M_j, D(M_j))} where I represents said identity of the closest reference model in the corresponding M_j^th subset of said M subsets and D(M_j) represents said computed distance between said closest reference model in the corresponding M_j^th subset and said temporary biometric model.

26. The program storage device of claim 22, further comprising instructions for performing the step of ranking the reference models in each corresponding one of said M subsets in accordance with the closeness of said reference models to the temporary biometric model, wherein said pattern of parameters further comprises ranking parameters corresponding to said ranking of said reference models in each of said M subsets.

27. The program storage device of claim 26, wherein said pattern of parameters comprises a pattern {[I(M₁, R₁), . . . I(M₁, R_i)], . . . , [I(M_j, R₁), . . . I(M_j, R_i)]}, where I is said identity of an R_i^th closest reference model in the corresponding M_j^th subset of said plurality of M subsets.

28. The program storage device of claim 26, wherein said pattern of parameters further comprises distance parameters corresponding to said computed distance between said identified and ranked reference models and said temporary biometric model.

29. The program storage device of claim 28, wherein said pattern of parameters comprises a pattern {[I(M₁, R₁, D(M₁, R₁)), . . . I(M₁, R_i, D(M₁, R_i))], . . . , [I(M_j, R₁, D(M_j, R₁)), . . . , I(M_j, R_i, D(M_j, R_i))]}, where I represents said identity of the R_i^th closest reference model in the corresponding M_j^th subset of said plurality of M subsets and D represents said corresponding computed distance.