Using statistical learning methods to fuse parameter estimates

• US 8,175,851 B1
• Filed: 09/24/2009
• Issued: 05/08/2012
• Est. Priority Date: 09/24/2009
• Status: Active Grant

• Current Assignee: Engility LLC (Science Applications International Corporation)
• Original Assignee: L-3 Services Incorporated (L3Harris Technologies, Inc.)
• Inventors: Francis, James Covosso, McGann, David Delaney
• Primary Examiner(s): Tarcza, Thomas
• Assistant Examiner(s): Mull, Fred H
• Application Number:
  US12/566,480
• Time in Patent Office: 957 Days
• Field of Search: 342/451, 702/189, 702/1, 702/127, 702/199, 708/100, 708/105, 708/131, 708/160, 708/200
• US Class Current: 702/199
• CPC Class Codes: G01S 3/043 Receivers
• CPC Class Codes: G01S 3/46 using antennas spaced apart and measuring phase or time difference between signals therefrom, i.e. path-difference systems
• CPC Class Codes: G01S 3/74 Multi-channel systems specially adapted for direction-finding, i.e. having a single antenna system capable of giving simultaneous indications of the directions of different signals systems in which the directions of different signals are determined sequentially and displayed simultaneously G01S3/04, G01S3/14

Abstract

A preferred estimate of a given parameter is made by processing a set of data points representing estimates of the given parameter, such as the angle of arrival of a target signal. The process includes the steps of: (a) obtaining a set of data points representing estimates of a given parameter; (b) using a computer to process said data points with an unsupervised clustering algorithm to select data points for use in making a preferred estimate of the given parameter; and (c) using the selected data points to make a preferred estimate of the given parameter.

Claims

1. A method of making a preferred estimate of a given parameter by processing a set of data points representing estimates of the given parameter, comprising the steps of:

(a) obtaining a set of data points representing estimates of a given parameter;

(b) using a computer to process said data points with an unsupervised clustering algorithm to select data points for use in making a preferred estimate of the given parameter; and

(c) using the selected data points to make a preferred estimate of the given parameter;

wherein step (b) comprises the steps of

(d) organizing the set of data points representing estimates of the given parameter into different clusters of said data points by assigning the respective data points to their closest cluster;

(e) after all of the data points have been assigned, averaging all of the data points in each cluster to calculate a new data point at the center of each cluster;

(f) when at least some of the most recently calculated data points do not converge into a preferred cluster having predetermined characteristics, repeating steps (d) and (e) with the most recently calculated data points until at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics; and

(g) when at least some, but not all, of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics selecting the data points of the preferred cluster for use in making the preferred estimate of the given parameter.

2. A method according to claim 1, wherein the preferred cluster is the cluster with the largest number of data points that are less than a threshold dimension from the center of the cluster.

3. A method according to claim 1, wherein step (c) comprises making the preferred estimate of the given parameter by determining a weighted average of the selected data points in the preferred cluster, with individual data point being weighted by a likelihood ratio of the probability of the estimate represented by the individual data point in relation to another possible estimate.

4. A method according to claim 3, wherein when the given parameter is the angle of arrival (AOA) of a target signal received by an array of antenna elements and the set of data points is derived from samples of the received signal, the likelihood ratio is either:

(i) E_global/(E_total−

E_global) when no local maximum exists after removing the effects of an antenna response that is dependent on the AOA of the received signal, wherein E_total is the total energy of the samples of the target signal from which the set of data points is derived, and E_global is the energy of the global maximum of the received signal after removing the effects of an antenna response that is dependent on the AOA;

or

(ii) (E_total−

E_local)/(E_total−

E_global), wherein when a local maximum exists after removing the effects of an antenna response that is dependent on the AOA of the received signal, wherein E_local is the energy of the local maximum signal after removing the effects of an antenna response that is dependent on the AOA.

5. A method according to claim 1, wherein the unsupervised clustering algorithm is a k-means algorithm.

6. A method according to claim 1, wherein the given parameter is the angle of arrival of a target signal detected by an array of antenna elements.

7. A method of making a preferred estimate of a given parameter by processing a set of data points representing estimates of the given parameter, comprising the steps of:

(a) obtaining a set of data points representing estimates of a given parameter;

(b) using a computer to process said data points with an unsupervised clustering algorithm to select data points for use in making a preferred estimate of the given parameter; and

(c) using the selected data points to make a preferred estimate of the given parameter;

wherein step (b) comprises the steps of

(d) organizing the set of data points representing estimates of the given parameter into different clusters of said data points by assigning the respective data points to their closest cluster;

(e) after all of the data points have been assigned, averaging all of the data points in each cluster to calculate a new data point at the center of each cluster;

(f) when at least some of the most recently calculated data points do not converge into a preferred cluster having predetermined characteristics, repeating steps (d) and (e) with the most recently calculated data points until at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics; and

(g) when at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics selecting the data points of the preferred cluster for use in making the preferred estimate of the given parameter;

wherein the preferred cluster is the cluster with the largest number of data points that are less than a threshold dimension from the center of the cluster; and

wherein when the given parameter is the angle of arrival (AOA) of, a target signal detected by an array of antenna elements, and the data points of the two clusters having the largest numbers of data points are less than the threshold dimension from the center of the particular cluster, the respective data points of both clusters are selected for making separate preferred estimates of angles of arrival of different target signals.

8. A system for making a preferred estimate of a given parameter by processing a set of data points representing estimates of the given parameter, comprising:

a computer adapted for obtaining a set of data points representing estimates of a given parameter by processing samples of a signal having the given parameter, also adapted for processing said data points with an unsupervised clustering algorithm to select data points for use in making a preferred estimate of the given parameter and further adapted for using the selected data points to make a preferred estimate of the given parameter;

wherein the computer is adapted for said processing of the data points by performing the routines of;

(a) organizing the set of data points representing estimates of the given parameter into different clusters of said data points by assigning the respective data points to their closest cluster;

(b) after all of the data points have been assigned, averaging all of the data points in each cluster to calculate a new data point at the center of each cluster;

(c) when at least some of the most recently calculated data points do not converge into a preferred cluster having predetermined characteristics, repeating steps (a) and (b) with the most recently calculated data points until at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics; and

(d) when at least some, but not all, of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics selecting the data points of the preferred cluster for use in making the preferred estimate of the given parameter.

9. A system according to claim 8, wherein the preferred cluster is the cluster with the largest number of data points that are less than a threshold dimension from the center of the cluster.

10. A system, according to claim 8, wherein the means for using the selected data points is adapted for making the preferred estimate of the given parameter by determining a weighted average of the selected data points in the preferred cluster, with individual data point being weighted by a likelihood ratio of the probability of the estimate represented by the individual data point in relation to another possible estimate.

11. A system according to claim 10, wherein when the given parameter is the angle of arrival (AOA) of a target signal received by an array of antenna elements and the set of data points is derived from samples of the received signal, the likelihood ratio is either:

E_global/(E_total−

E_global) when no local maximum exists after removing the effects of an antenna response that is dependent on the AOA of the received signal, wherein E_total is the total energy of the samples of the target signal from which the set of data points is derived, and E_global is the energy of the global maximum of the received signal after removing the effects of an antenna response that is dependent on the AOA;

or

(E_total−

E_local)/(E_total−

E_global), wherein when a local maximum exists after removing the effects of an antenna response that is dependent on the AOA of the received signal, wherein E_local is the energy of the local maximum signal after removing the effects of an antenna response that is dependent on the AOA.

12. A system according to claim 8, wherein the unsupervised clustering algorithm is a k-means algorithm.

13. A system according to claim 8, wherein the given parameter is the angle of arrival of a target signal detected by an array of antenna elements.

14. A system for making a preferred estimate of a given parameter by processing a set of data points representing estimates of the given parameter, comprising:

a computer adapted for obtaining a set of data points representing estimates of a given parameter by processing samples of a signal having, the given parameter, also a adapted for processing said data points with an unsupervised clustering algorithm to select data points for use in making a preferred estimate of the given parameter and further adapted for using the selected data points to make a preferred estimate of the given parameter;

wherein the computer is adapted for said processing of the data points by performing the routines of

(a) organizing the set of data points representing estimates of the given parameter into different clusters of said data points by assigning the respective data points to their closest cluster;

(b) after all of the data points have been assigned, averaging all of the data points in each cluster to calculate a new data point at the center of each cluster;

(c) when at least some of the most recently calculated data points do not converge into a preferred cluster having predetermined characteristics, repeating steps (a) and (b) with the most recently calculated data points until at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics; and

(d) when at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics selecting the data points of the preferred cluster for use in, making the preferred estimate of the given parameter; and

wherein when the given parameter is the angle of arrival of a target signal detected by an array of antenna elements, and the data points of the two clusters having the largest numbers of data points are less than the threshold dimension from the center of the particular cluster, the respective data points of both clusters are selected for making separate preferred estimates of angles of arrival of different target signals.

15. A nontransitory computer readable storage medium containing computer executable program instructions for use with a computer in a system for making a preferred estimate of a given parameter by processing a set of data points representing estimates of the given parameter, wherein the system comprises:

means for obtaining a sot of data points representing estimates of a given parameter;

a computer for processing said data points to select data points for use in making a preferred estimate of the given parameter; and

means for using the selected data points to make the preferred estimate of the given parameter,

wherein the computer readable storage medium contains computer executable to program instructions for causing the computer to process said data points with an unsupervised clustering algorithm to select said data points for use in making the preferred estimate of the given parameter; and

for causing the computer to select said data points for use in making the preferred estimate of the given parameter by performing the routines of;

(a) organizing the set of data points representing estimates of the given parameter into different clusters of said data points by assigning the respective data points to their closest cluster;

(b) after all of the data points have been assigned, averaging all of the data points in each cluster to calculate a new data point at the center of each cluster;

(c) when at least some of the most recently calculated data points do not converge into a preferred cluster having predetermined characteristics, repeating steps (a) and (b) with the most recently calculated data points until at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics; and

(d) when at least some, but not all, of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics selecting the data points of the preferred cluster for use in making the preferred estimate of the given parameter.

16. A nontransitory computer readable storage medium according to claim 15, wherein the preferred cluster is the cluster with the largest number of data points that are less than a threshold dimension from the center, of the cluster.

17. A nontransitory computer readable storage medium according to claim 15, further containing computer executable program instructions for causing a computer to process the selected data points to make the preferred estimate of the given parameter by determining a weighted average of the selected data points in the preferred cluster, with individual data point being weighted by a likelihood ratio of the probability of the estimate represented by the individual data point in relation to another possible estimate.

18. A nontransitory, computer readable storage medium according to claim 17, wherein when the given parameter is the angle of arrival (AOA) of a target signal received by an array of antenna element'"'"'s and the set of data points is derived from samples of the received signal, the likelihood ratio is either:

(i) E_global/(E_total−

E_global) when no local maximum exists after removing the effects of an antenna response that is dependent on the AOA of the received signal, wherein E_total is the total energy of the samples of the target signal from which the set of data points is derived, and E_global is the energy of the global maximum of the received signal after removing the effects of an antenna response that is dependent on the AOA;

or

(ii) (E_total−

E_local)/(E_total−

E_global), wherein when a local maximum exists after removing the effects of an antenna response that is dependent on the AOA of the received signal, wherein E_local is the energy of the local maximum signal after removing the effects of an antenna response that is dependent on the AOA.

19. A nontransitory computer readable storage medium according to claim 15, wherein the unsupervised clustering algorithm is a k-means algorithm.

20. A nontransitory computer readable storage medium according to claim 15, wherein the given parameter is the angle of arrival of a target signal detected by an array of antenna elements.

21. A nontransitory computer readable storage medium containing computer executable program instructions for use with a computer in a system for making a preferred estimate of a given parameter by processing a set of data points representing estimates of the given parameter, wherein the system comprises:

means for obtaining a set of data points representing estimates of a given parameter;

a computer for processing said data points to select data points for use in making a preferred estimate of the given parameter; and

means for using the selected data points to make the preferred estimate of the given parameter,

wherein the computer readable storage medium contains computer executable program instructions for causing the computer, to process said data points with an unsupervised clustering algorithm to select said data points for use in making the preferred estimate of the given parameter; and

for causing the computer to select said data points for use in making the preferred estimate of the given parameter by performing the routines of;

(a) organizing the set of data points representing estimates of the given parameter into different clusters of said data points by assigning the respective data points to their closest cluster;

(b) after all of the data points have been assigned, averaging all of the data points in each cluster to calculate a new data point at the center of each cluster;

(c) when at least some of the most recently calculated data points do not converge into a preferred cluster having predetermined characteristics, repeating steps (a) and (b) with the most recently calculated data points until at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics; and

(d) when at least some of the most recently calculated data points converge into a preferred cluster having the predetermined characteristics selecting the data points of the preferred cluster for use in making the preferred estimate of the given parameter;

wherein the preferred cluster is the cluster with the largest number of data points that are less than a threshold dimension from the center of the cluster; and

wherein when the given parameter is the angle of arrival of a target signal detected by an array of antenna elements, and the data points of the two clusters having the largest numbers of data points are less than the threshold dimension from the center of the particular cluster, the respective data points of both clusters are selected for making separate preferred estimates of angles of arrival of different target signals.