METHOD AND APPARATUS FOR COMBINING OUTPUTS OF MULTIPLE SYSTEMS

• US 20070211923A1
• Filed: 03/13/2006
• Published: 09/13/2007
• Est. Priority Date: 03/13/2006
• Status: Active Grant

• Current Assignee: Motorola Solutions, Inc.
• Original Assignee: Motorola, Inc. (Motorola Solutions, Inc.)
• Inventors: Kuhlman, Douglas
• Granted Patent: US 7,565,265 B2
• Time in Patent Office: Days
• US Class Current: 382/116
• CPC Class Codes: G06F 18/256 of results relating to different input data, e.g. multimodal recognition

Abstract

Automatic recognition systems (100) includes multiple automatic recognition subsystems (102, 104, 106) that are cable of producing estimates of the probability that a subject matches a particular identity and a probability estimate combiner (108) that receives estimates from the multiple automatic recognition subsystems (102, 104, 106). The probability estimate combiner (108) has a number of properties which allow good use to be made of the individual estimates.

Claims

1. A probability estimate combiner comprising:

a processing unit;

a first input coupled to said processing unit;

a second input coupled to said processing unit;

an output coupled to said processing unit;

wherein said processing unit is adapted to execute a function C that has;

a first property by which a quantity C(X, X) is further from a predetermined value X₀ than X, for all values of X in a predetermined domain of X not equal to X₀ and not equal to a set of bounds of the predetermined domain, wherein X₀ is an interior point in the predetermined domain that corresponds to equal probability that a result is true and false;

a second property by which C(X, X₀) is equal to X; and

a third property by which a quantity C(X₀−

Δ

X, X₀+Δ

X) is equal to X₀, where Δ

X is an increment

wherein, the processing unit outputs, at the output, a combined probability estimate that is obtained by evaluating the function C.

2. A method of performing automatic identification comprising:

operating a first automatic identification system to obtain a first estimate of a probability that a subject has a particular identity;

operating a second automatic identification system to obtain a second estimate of the probability that the subject has the particular identity;

inputting the first estimate and the second estimate of the probability to a function C of at least the first estimate and the second estimate that has;

a first property by which a quantity C(X, X) is further from a first predetermined value X₀ than X, for all values of X in a predetermined domain of X not equal to X₀ and not equal to a set of bounds of the predetermined domain, wherein X₀ is an interior point in the predetermined domain that corresponds to equal probability that the subject does and does not have the particular identity;

a second property by which C(X, X₀) is equal to X; and

evaluating the function C to obtain a combined estimate of the probability that the subject has the particular identity.

3. The method of performing automatic identification according to claim 2 wherein inputting the first estimate and the second estimate of the probability to the function C comprises:

inputting the first estimate and the second estimate of the probability to a function that has;

a third property by which a quantity C(X₀−

Δ

X, X₀+Δ

X) is equal to X₀, where Δ

X is an increment.

4. The method of performing automatic identification according to claim 2 wherein inputting the first estimate and the second estimate of the probability to the function C of comprises:

inputting the first estimate and the second estimate of the probability to a function that has the commutative property and the associative property.

5. The method of performing automatic identification according to claim 2 wherein inputting the first estimate and the second estimate of the probability to the function C comprises:

inputting the first estimate and the second estimate of the probability to a function that has a third property that, at least over the predetermined domain, a range of the function is equal to the predetermined domain.

6. The method of performing automatic identification according to claim 2 wherein inputting the first estimate and the second estimate of the probability to the function C comprises:

inputting the first estimate and the second estimate of the probability to a function selected from the group consisting of;

$C<br><br>\left(x,y\right)=\frac{2<br><br>\mathrm{xy}}{1+\left(2<br><br>x-1\right)<br><br>\left(2<br><br>y-1\right)},<br><br>C<br><br>\left(x,y\right)=\frac{x+y}{1+\mathrm{xy}},\mathrm{and}$$C<br><br>\left(x,y\right)=-\frac{1-\sqrt{1+4<br><br>{f<br><br>\left(x,y\right)}^2}}{2<br><br>f<br><br>\left(x,y\right)}$$\mathrm{in}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{which}$$f<br><br>\left(x,y\right)=\frac{\left(x+y\right)<br><br>\left(1-\mathrm{xy}\right)}{\left(1-x^2\right)<br><br>\left(1-y^2\right)}.$

7. The method of performing automatic identification according to claim 2 wherein:

obtaining the first estimate of the probability that the subject has the particular identity comprises;

operating a statistical pattern recognition system.

8. The method of performing automatic identification according to claim 7 wherein:

operating the statistical pattern recognition system comprises;

operating a biometric statistical pattern recognition system.

9. A system comprising:

a first automatic identification subsystem adapted to take data from a subject and to output a first estimate of a probability that the subject has a particular identity;

a second automatic identification subsystem adapted to take data from the subject and to output a second estimate of the probability that the subject has the particular identity;

a function evaluator adapted to receive the first estimate and the second estimate and to evaluate a function C of the first estimate and the second estimate wherein the function C has;

a first property by which a quantity C(X, X) is further from a first predetermined value X₀ than X, for all values of X in a predetermined domain of X not equal to X₀ and not equal to a set of bounds of the predetermined domain, wherein X₀ is an interior point in the predetermined domain that corresponds to equal probability that the subject does and does not have the particular identity;

a second property by which C(X, X₀) is equal to X; and

wherein, the function evaluator is further adapted to output a combined estimate of the probability that the subject has the particular identity, that is obtained by evaluating the function C.

10. The system according to claim 9 wherein the function evaluator is adapted to evaluate the function that has:

a third property by which a quantity C(X₀−

Δ

X, X₀+Δ

X) is equal to X₀, where Δ

X is an increment.

11. The system according to claim 9 wherein the function evaluator is adapted to evaluate the function that has:

the commutative property and the associative property.

12. The system according to claim 9 wherein the function evaluator is adapted to evaluate the function that has:

a third property that, at least over the predetermined domain, a range of the function is equal to the predetermined domain.

13. The system according to claim 9 wherein the function evaluator is adapted to evaluate a function selected from the group consisting of:

14. The system according to claim 9 wherein the first automatic identification subsystem comprises:

a statistical pattern recognition subsystem.

15. The system according to claim 14 wherein the statistical pattern recognition subsystem comprises:

a biometric statistical pattern recognition system.

16. A computer readable medium storing programming instructions for performing automatic identification including programming instructions for:

operating a first automatic identification system to obtain a first estimate of a probability that a subject has a particular identity;

operating a second automatic identification system to obtain a second estimate of the probability that the subject has the particular identity;

inputting the first estimate and the second estimate of the probability to a function C of at least the first estimate and the second estimate that has;

a first property by which a quantity C(X, X) is further from a first predetermined value X₀ than X, for all values of X in a predetermined domain of X not equal to X₀ and not equal to a set of bounds of the predetermined domain, wherein X₀ is an interior point in the predetermined domain that corresponds to equal probability that the subject does and does not have the particular identity;

a second property by which C(X, X₀) is equal to X; and

evaluating the function C to obtain a combined estimate of the probability that the subject has the particular identity.

17. The computer readable media according to claim 16 wherein the programming instructions for inputting the first estimate and the second estimate of the probability to the function C comprise programming instructions for:

inputting the first estimate and the second estimate of the probability to a function that has;

a third property by which a quantity C(X₀−

Δ

X, X₀+Δ

X) is equal to X₀, where Δ

X is an increment.

18. The computer readable medium according to claim 16 wherein the programming instructions for inputting the first estimate and the second estimate of the probability to the function C of comprise programming instructions for:

inputting the first estimate and the second estimate of the probability to a function that has the commutative property and the associative property.

19. The computer readable medium according to claim 16 wherein the programming instructions for inputting the first estimate and the second estimate of the probability to the function C comprise programming instructions for:

inputting the first estimate and the second estimate of the probability to a function that has a third property that, at least over the predetermined domain, a range of the function is equal to the predetermined domain.

20. The computer readable medium according to claim 16 wherein the programming instructions for inputting the first estimate and the second estimate of the probability to the function C comprise programming instructions for:

inputting the first estimate and the second estimate of the probability to a function selected from the group consisting of;

$C<br><br>\left(x,y\right)=\frac{2<br><br>\mathrm{xy}}{1+\left(2<br><br>x-1\right)<br><br>\left(2<br><br>y-1\right)},<br><br>C<br><br>\left(x,y\right)=\frac{x+y}{1+\mathrm{xy}},\mathrm{and}$$C<br><br>\left(x,y\right)=-\frac{1-\sqrt{1+4<br><br>{f<br><br>\left(x,y\right)}^2}}{2<br><br>f<br><br>\left(x,y\right)}$$\mathrm{in}<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{which}$$f<br><br>\left(x,y\right)=\frac{\left(x+y\right)<br><br>\left(1-\mathrm{xy}\right)}{\left(1-x^2\right)<br><br>\left(1-y^2\right)}.$

21. The computer readable medium according to claim 16 wherein the programming instructions for:

operating the first automatic identification system to obtain the first estimate of the probability that the subject has the particular identity comprise programming instructions for;

operating a statistical pattern recognition system.

22. The computer readable medium according to claim 21 wherein the programming instructions for:

operating the statistical pattern recognition system comprise programming instructions for;

operating a biometric statistical pattern recognition system.

23. A method of performing automatic identification comprising:

receiving a first estimate of a probability that a subject has a particular identity from a first automatic identification system;

receiving a second estimate of the probability that the subject has the particular identity from a second automatic identification system;

inputting the first estimate and the second estimate of the probability to a function C of at least the first estimate and the second estimate that has;

a first property by which a quantity C(X, X) is further from a first predetermined value X₀ than X, for all values of X in a predetermined domain of X not equal to X₀ and not equal to a set of bounds of the predetermined domain, wherein X₀ is an interior point in the predetermined domain that corresponds to equal probability that the subject does and does not have the particular identity;

a second property by which C(X, X₀) is equal to X; and

evaluating the function C to obtain a combined estimate of the probability that the subject has the particular identity.