Biometric subject verification based on electrocardiographic signals

• US 8,924,736 B1
• Filed: 03/11/2013
• Issued: 12/30/2014
• Est. Priority Date: 03/11/2013
• Status: Active Grant

• Current Assignee: USA As Represented By The Administrator Of The NASA
• Original Assignee: The United States of America As Represented By The Secretary of Agriculture
• Inventors: Dusan, Sorin V., Jorgensen, Charles C.
• Primary Examiner(s): Rahman, Mohammad L
• Assistant Examiner(s): CHANG, KENNETH W
• Application Number:
  US13/793,998
• Time in Patent Office: 659 Days
• Field of Search: 726/19
• US Class Current: 713/186
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G06F 2218/12 Classification; Matching
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
• CPC Class Codes: G06V 40/15 Biometric patterns based on physiological signals, e.g. heartbeat, blood flow

Abstract

A method of authenticating or declining to authenticate an asserted identity of a candidate-person. In an enrollment phase, a reference PQRST heart action graph is provided or constructed from information obtained from a plurality of graphs that resemble each other for a known reference person, using a first graph comparison metric. In a verification phase, a candidate-person asserts his/her identity and presents a plurality of his/her heart cycle graphs. If a sufficient number of the candidate-person'"'"'s measured graphs resemble each other, a representative composite graph is constructed from the candidate-person'"'"'s graphs and is compared with a composite reference graph, for the person whose identity is asserted, using a second graph comparison metric. When the second metric value lies in a selected range, the candidate-person'"'"'s assertion of identity is accepted.

Claims

1. A method of authenticating or declining to authenticate an asserted identity of a candidate person (c-p), the method comprising:

(i) initializing a count index c (c(initial)=0);

(ii) providing or receiving a sequence of N reference heart electrical action cycles, numbered n=1, . . . , N, of a known reference person, and forming a graph v(t;

n) representing each provided heart cycle for the reference person;

(iii) randomly choosing K heart cycle graphs, numbered k=1, 2, . . . , K (3≦

K≦

N) of the provided reference heart cycles, comparing each chosen heart cycle graph v(t;

k) with each of the other chosen heart cycle graphs, and identifying a pair of graphs, v(t;

k1) and v(t;

k2), for which ρ

1(k1;

k2) is a maximum value, where ρ

1 is a first numerical-valued, graph comparison metric that compares two heart cycle graphs, v(t;

k′

) and v(t;

k″

), and provides a first numerical metric value ρ

1(k′

;

k″

) that is a measure of how closely the two graphs resemble each other;

(iv) forming a composite heart cycle graph, denoted v(t;

k12), using graph values associated with the graphs v(t;

k1) and v(t;

k2);

(v) determining when at least a first selected fraction f1 (0.5≦

f1<

1) of the N heart cycle graphs resemble the composite graph, according to a first condition ρ

1(k12;

n)−

(thr1)=D1, with D1 being always non-negative or always non-positive, where (thr1) is a selected first metric threshold value;

(vi) when at least the selected fraction f1 of the N heart cycle graphs resemble each other according to the first condition, creating or providing a reference heart cycle graph v(t;

ref)=v(t;

k12);

for the reference person;

(vii) when at least the fraction f1 of the N heart cycle graphs do not resemble each other, incrementing the count index (c→

c+1) to provide an incremented count index;

(viii) when the incremented count index c is not greater than a selected maximum index value c(max), returning to step (iii) and repeating steps (iii)-(viii) at least once; and

(ix) when the incremented count index satisfies c>

c(max), interpreting this condition as indicating that it is not possible to create a reference heart cycle graph for the reference person.

2. The method of claim 1, wherein said process of comparing two heart cycle graphs comprises:

(iii-1) forming a correlation of said heart cycle graphs, v(t;

k=k1) and v(t;

k=k2), according to a relation

3. The method of claim 2, further comprising:

(iii-4) when ρ

1 satisfies |ρ

1(k1;

k2)|−

(thr1)=D1<

0, for said threshold metric value (thr), interpreting this condition as indicating that the heart cycle graphs v(t;

k1) and v(t;

k2) do not resemble each other.

4. The method of claim 1, wherein said process of comparing two heart cycle graphs comprises

(iii-1) forming said first metric as a weighted sum of differences of said reference heart cycle graphs, v(t;

k=k1) and v(t;

k=k2), according to a relation

5. The method of claim 4, further comprising:

(iii-3) where ρ

1(k1;

k2) satisfies |ρ

1(k1;

k2)|−

(thr1′

)>

0, for said threshold metric value (thr1′

), interpreting this condition as indicating that the heart cycle graphs v(t;

k1) and v(t;

k2) do not resemble each other.

6. The method of claim 1, further comprising providing said reference heart cycle graph by a process comprising:

defining a maximum first metric value ρ

1(k1,k2) as a value equal to max<sub>k′

, k″</sub>

{ρ

1(k′

,k″

)}; and

defining said reference heart cycle graph v(t;

ref) as Avg_k1,k2{v(t;

k1), v(t;

k2)}.

7. The method of claim 6, further comprising choosing said average function Avg{v(t;

k1), v(t;

k2)} from the group consisting of;

Avg{v(t;

k1),v(t;

k2)}={v(t;

k1)+v(t;

k2)}/2, 

 

(i)

Avg{v(t;

k1),v(t;

k2)}=sgn{v(t;

k1)+v(t;

k2)}|v(t;

k1)·

v(t;

k2)|^1/2, and 

 

(ii)

Avg{v(t;

k1),v(t;

k2)}=v(t;

k1)+v(t;

k2)−

sgn{v(t;

k1)+v(t;

k2)}·

|v(t;

k1)·

v(t;

k2)|^1/2. 

 

(iii)

8. The method of claim 1, further comprising:

passing said sequence of N reference graphs through a digital filter that (i) suppresses or removes all frequency contributions to said N reference graphs below a first threshold frequency f(low;

1) and (ii) removes or suppresses all frequency contributions to said N reference graphs above a second threshold frequency f(high;

2) that is greater than the first threshold frequency f(low;

1).

9. The method of claim 8, further comprising choosing said first threshold frequency f(low;

1) to be no greater than about 2 Hz.

10. The method of claim 8, further comprising choosing said second threshold frequency f(high;

2) to be no less than about 42 Hz.

11. The method of claim 1, further comprising:

(x) initializing a count index c′

(c′

(initial)=0);

(xi) providing or receiving a sequence of N′

heart electrical action cycles, numbered n′

=1, . . . , N′

, from a candidate-person (c-p), and forming a graph v′

(t;

n′

;

c-p) representing each heart cycle for the candidate-person;

(xii) randomly choosing K′

c-p heart cycle graphs, numbered k′

=1, 2, . . . , K′

(3≧

K′

≧

N′

) of the provided c-p heart cycle graphs, comparing each chosen c-p heart cycle graph v′

(t;

k′

;

c-p) with each of the other chosen c-p heart cycle graphs, and identifying a pair of heart cycle graphs, v′

(t;

k1′

;

c-p) and v′

(t;

k2′

;

c-p), for which a value ρ

1(k1′

;

k2′

;

c-p) is a maximum value;

(xiii) forming a c-p composite heart cycle graph, denoted v′

(t;

k12′

;

c-p), using data associated with the graphs v′

(t;

k1′

;

c-p) and v′

(t;

k2′

;

c-p);

(xiv) determining when at least a second selected fraction f2 (0.5≦

f2<

1) of the N′

c-p heart cycle graphs resemble the c-p composite graph, according to a second condition ρ

1(k12′

;

n′

)−

(thr1′

)=D1′

, where (thr1′

) is a second selected first metric threshold value and D1′

is always non-negative or is always non-positive;

(xv) when at least the second selected fraction f2 of the N′

c-p heart cycle graphs resemble each other, providing a representative heart cycle graph v′

(t;

rep)=v′

(t;

k12′

;

c-p);

(xvi) comparing the representative graph v′

(t;

rep)=v′

(t;

k12′

;

c-p) for the candidate-person with said reference graph v(t;

ref)=v(t;

k12) for said reference person by comparing two graphs, v′

(t;

rep) and v(t;

ref), using a second, numerical valued, graph comparison metric ρ

2 that compares two graphs and provides a numerical second metric value ρ

2(rep;

ref) that is a measure of how closely the two graphs resemble each other;

(xvii) when ρ

2(rep;

ref) satisfies a third condition ρ

2(rep;

ref)−

(thr2)=D2, where (thr2) is a selected second metric threshold value and D2 is always non-negative or is always non-positive, interpreting satisfaction of the third condition as indicating that the candidate-person is likely to be said reference person;

(xviii) when ρ

2(rep;

ref) satisfies ρ

2(rep;

ref)−

(thr2)=D2″

, where D2″

is negative at least once and is positive at least once, interpreting this condition as indicating that the candidate-person is not likely to be said reference person;

(xix) when at least the second fraction f2 of the N′

heart cycle graphs do not resemble each other, incrementing the count index (c′

→

c′

+1), to provide an incremented count index c′

(incr);

(xx) when the incremented count index c′

(incr) is not greater than a selected maximum index value c′

(max), returning to step (xii) and repeating steps (xii)-(xx) at least once; and

(xxi) when the incremented count index satisfies c′

(incr)>

c′

(max), interpreting this condition as indicating that it is not possible to create a representative heart cycle graph for the candidate-person.

12. The method of claim 11, wherein said process of comparing two heart cycle graphs comprises:

(xvi-1) forming a correlation of said c-p heart cycle graphs, v′

(t;

k=k1;

c-p) and v′

(t;

k=k2;

c-p), according to a relation

13. The method of claim 11, further comprising:

(xiii-4) where ρ

2 satisfies ρ

2(k1;

k2)−

(thr2)=D2<

0, for said threshold metric value (thr2), interpreting this condition as indicating that the heart cycle graphs v′

(t;

k1;

c-p) and v′

(t;

k2;

c-p) do not resemble each other.

14. The method of claim 11, wherein said process of comparing two heart cycle graphs comprises

(xvi-1) forming said second metric as a weighted sum of differences of said c-p heart cycle graphs, v′

(t;

k=k1;

c-p) and v′

(t;

k=k2;

c-p), according to a relation

15. The method of claim 14, further comprising:

(xvi-3) when ρ

2 satisfies ρ

2(k1;

k2)−

(thr2)>

0, for said threshold metric value (thr2), interpreting this condition as indicating that said heart cycle graphs v′

(t;

k1;

c-p) and v′

(t;

k2;

c-p) do not resemble each other.

16. The method of claim 11, further comprising providing said representative c-p heart cycle graph by a process comprising:

defining a maximum first metric value ρ

1(k1,k2) as a value equal to max<sub>k′

,k″</sub>

{ρ

1(k′

,k″

)}; and

defining said representative c-p heart cycle graph v(t;

rep) as Avg_k1,k2{v(t;

k1), v(t;

k2)}, where Avg_k1,k2{v′

(t;

k1′

;

c-p), v′

(t;

k2′

;

c-p)} is an average value for said heart cycle graphs v′

(t;

k1′

c-p) and v′

(t;

k2′

;

c-p).

17. The method of claim 16, further comprising choosing said average function Avg{v(t;

k1′

;

c-p), v(t;

k2′

;

c-p)} from the group consisting of;

Avg{v′

(t;

k′

1;

c-p),v′

(t;

k2′

;

c-p)}={v′

(t;

k1′

;

c-p)+v′

(t;

k2′

;

c-p)}/2, 

 

(i)

Avg{v′

(t;

k1;

c-p),v′

(t;

k2′

;

c-p)}=sgn{v′

(t;

k1′

;

c-p)+v′

(t;

k2′

;

c-p)}|v′

(t;

k1′

;

c-p)·

v′

(t;

k2′

;

c-p)|^1/2, and 

 

(ii)

Avg{v′

(t;

k1′

;

c-p),v′

(t;

k2′

;

c-p)}=v′

(t;

k1′

;

c-p)+v′

(t;

k2′

;

c-p)−

sgn{v′

(t;

k1;

c-p)+v′

(t;

k2′

;

c-p)}·

|v′

(t;

k1′

;

c-p)·

v′

(t;

k2′

;

c-p)|^1/2. 

 

(iii)

18. The method of claim 11, further comprising:

estimating a heart beat rate HR, expressed in beats per minute (bpm) or an equivalent time rate for a heart cycle signature, for said reference person and for said candidate-person, where the heart electrical cycle signature comprises a first time interval plus a second time interval that does not overlap the first time interval except possibly at interval end points, where the first time interval ends and the second time interval for said reference person begins approximately at a central point, t=t(acp;

ref) and the first time interval and the second time interval for said c-p person begins approximately at a central point, t=t(acp;

c-p);

providing a first normalization factor α

1, relating a ratio of a length of the first time interval for the HR for said reference person to a first reference length of the first time interval for HR=60 bpm or the equivalent time rate;

providing a second normalization factor α

2, relating a ratio of a length of the first time interval for the HR for said c-p person to a second reference length of the first time interval for HR=60 bpm or the equivalent time rate;

providing a third normalization factor β

1, relating a ratio of a length of the second time interval for the HR for said reference person to a third reference length of the second time interval for HR=60 bpm or the equivalent time rate;

providing a second normalization factor β

2, relating a ratio of a length of the second time interval for the HR for said candidate-person to a fourth reference length of the second time interval for HR=60 bpm or the equivalent time rate;

replacing said reference graph v(t;

ref) by a corrected or modified reference graph that is defined as

v(t;

ref)_corr=v{t(acp);

ref)−

α

1(t(acp);

ref)−

t;

ref)} (t≦

t(acp);

ref)).

v(t;

ref)_corr=v{t(acp);

ref)+β

1(t−

t(acp);

ref);

ref)} (t(acp);

ref)≦

t); and

replacing said representative graph v′

(t;

rep) by a corrected or modified representative graph that is defined as

v′

(t;

rep)_corr=v′

{t(acp);

c-p)−

α

2(t(acp);

c-p)−

t;

rep)} (t≦

t(acp);

c-p)),

v′

(t;

rep)_corr=v′

{t(acp);

c-p)+β

2(t−

t(acp);

c-p);

rep)} (t(acp);

c-p)≦

t).

19. The method of claim 18, further comprising defining said first normalization factor α

=α

1 or α

=α

2 to be α

={195.36−

0.384(HR)}/{195.36−

0.384(60)}.

20. The method of claim 18, further comprising defining said first normalization factor β

=β

1 or β

2 to be β

={HR/60}^r, where r is a selected fraction in a range ⅓

≦

r≦

½

.

21. The method of claim 11, further comprising:

passing said K′

c-p heart cycle graphs through a digital filter that (i) suppresses or removes all frequency contributions to said K′

c-p heart cycle graphs below a third threshold frequency f(low;

3) and (ii) removes or suppresses all frequency contributions to said K′

c-p heart cycle graphs above a fourth threshold frequency f(high;

4) that is greater than the third threshold frequency f(low;

3).

22. The method of claim 11, further comprising choosing said third threshold frequency f(low;

3) to be no greater than about 2 Hz.

23. The method of claim 11, further comprising choosing said fourth threshold frequency f(high;

4) to be no less than about 42 Hz.

24. A method of authenticating or declining to authenticate an asserted identity of a candidate-person, the method comprising:

(i) initiating a count index c′

(c′

(initial)=0);

(ii) receiving an assertion, by a candidate-person (c-p), that the candidate-person is identical with a specified reference person;

(iii) providing a sequence of N′

heart electrical action cycles of the candidate-person, and forming a c-p heart cycle graph representing each provided c-p heart cycle;

(iv) randomly choosing K′

graphs, numbered k′

=1, 2, . . . , K′

(3≦

K′

≦

N) of the provided c-p heart cycles, and comparing each chosen c-p heart cycle graph with each of the other chosen c-p heart cycle graphs;

(v) determining whether a majority K′

_M of the K′

c-p heart cycle graphs resemble each other, according to a second selected numerical-valued, graph comparison metric ρ

2 that compares two heart cycle graphs, k′

=k″ and

k′

=k″

′

, and provides a second graph comparison metric value ρ

2(k″

;

k″

′

) for the comparison;

(vi) when the majority K′

_M of the K′

c-p heart cycle graphs resemble each other, providing a representative c-p heart cycle graph representing the majority K′

_M of the chosen c-p heart cycle graphs;

(vii) providing a reference heart cycle graph that corresponds to the identity asserted by the candidate-person;

(viii) comparing the representative c-p heart cycle graph with the reference heart cycle graph, using the second graph comparison metric ρ

2(rep;

ref), applied to the representative heart cycle graph and to the reference heart cycle graph;

(ix) when the value ρ

2(rep;

ref) provided by the second metric ρ

2 satisfies ρ

2(rep;

ref)−

ρ

2(thr2)=D2, where ρ

2(thr2) is a second selected threshold value and D2 is a real number that is always non-negative or that is always non-positive, interpreting this condition as indicating that the candidate-person is likely to be the reference person;

(x) when the value provided by the second metric ρ

2c does not satisfy the condition ρ

2(rep;

ref)−

ρ

2(thr2)=D2, with D2 always non-negative or D2 always non-positive, interpreting this condition as indicating that the candidate-person is not likely to be the reference person;

(xi) when the majority K′

_M of the N′

c-p heart cycle graphs do not resemble each other, incrementing the count index c′

(c′

→

c′

+1);

(xii) when the incremented count index c′

is no greater than a selected maximum index value c′

(max), returning to step (iv); and

(xiii) when the incremented count index c′

satisfies c′

>

c′

(max), interpreting this condition as indicating that it is not possible to create a representative heart cycle graph for the candidate-person.