Age compensation in biometric systems using time-interval, gender and age
A score indicating a likelihood that a first subject is the same as a second subject may be calibrated to compensate for aging of the first subject between samples of age-sensitive biometric characteristics. Age of the first subject obtained at a first sample time and age of the second subject obtained at a second sample time may be averaged, and an age approximation may be generated based on at least the age average and an interval between the first and second samples. The age approximation, the interval between the first and second sample times, and an obtained gender of the subject are used to calibrate the likelihood score.
1. A method comprising:
obtaining, by a server, a first sample of an age-sensitive biometric characteristic of an enrollee, the first sample collected from a first electronic device associated with the enrollee at a first sample time;
obtaining, by the server, an age of the enrollee;
obtaining, by the server, a second sample of an age-sensitive biometric characteristic, the second sample collected from a second electronic device associated with a caller at a second sample time after the first sample time;
obtaining, by the server, an age of the caller;
obtaining, by the server, a gender of at least one of the enrollee and the caller;
determining, by the server, an interval between the first sample time and the second sample time;
determining, by the server, an average age using the obtained age of the enrollee and the obtained age of the caller;
generating, by the server, an age approximation using the average age and the interval;
generating, by the server, a first likelihood score that the caller is the enrollee based on the age-sensitive characteristic of the first sample and the age-sensitive characteristic of the second sample;
calibrating, by the server, the first likelihood score based on the gender, the age approximation, and the interval, thereby generating a second likelihood score; and
authenticating, by the server, the caller when the second likelihood score satisfies a threshold.
2. The method according to claim 1 , wherein at least one of the age of the enrollee and the age of the caller is an age estimate, an age estimate of the enrollee being generated from the first sample and an age estimate of the caller being generated from the second sample.
3. The method according to claim 2 , wherein at least generating the age estimate of the caller is performed only when the interval is greater than a predetermined time threshold.
4. The method according to claim 2 , wherein generating the age estimate of the enrollee includes at least one of weighting and offsetting based on the obtained gender.
5. The method according to claim 2 , wherein generating the age estimate of the caller includes at least one of weighting and offsetting based on the obtained gender.
6. The method according to claim 1 , wherein the obtained gender is a gender estimate generated using information collected during the first sample time, the second sample time, or another time.
7. The method according to claim 1 , further comprising: using a calibrated likelihood score determination algorithm to adjust the first likelihood score for indicating whether the enrollee and the caller are the same person.
8. The device according to claim 7 , further comprising: using a deep neural network to determine the first likelihood score, the deep neural network being trained to discriminate between callers based at least on a statistical model produced using the first sample.
9. The method according to claim 7 , further comprising: updating a statistical model for the enrollee using the second sample when the second likelihood score exceeds a predetermined threshold.
10. The method according to claim 1 , wherein at least one of the first sample and the second sample is an audio voice sample.
11. The method according to claim 1 , wherein at least one of the first sample and the second sample is a facial image sample.
12. The method according to claim 1 , wherein generating the age approximation further includes at least one of weighting and offsetting of the obtained age of the caller based on the interval.
13. A caller recognition apparatus comprising:
receiving circuitry configured to obtain a first sample of at least an age-sensitive biometric characteristic collected from a first electronic device associated with an enrollee at a first sample time and to obtain a second sample of an age-sensitive biometric characteristic collected from a second electronic device associated with a caller at a second sample time after the first sample time; and
at least one processor configured to:
obtain an age of the enrollee,
obtain an age of the caller,
obtain a gender of at least one of the enrollee and the caller,
determine an interval between the first sample time and the second sample time,
determine an average age using the obtained age of the enrollee and the obtained age of the caller,
generate an age approximation using the average age and the interval,
generate a first likelihood score that the caller is the enrollee based on the age-sensitive characteristic of the first sample and the age-sensitive characteristic of the second sample;
calibrate the first likelihood score based on the gender, the age approximation, and the interval, thereby generating a second likelihood score; and
authenticate the caller when the second likelihood score satisfies a threshold.
14. The caller recognition apparatus according to claim 13 , wherein at least one of the age of the enrollee and the age of the caller is an age estimate, an age estimate of the enrollee being generated by the at least one processor from the first sample and an age estimate of the caller being generated by the at least one processor from the second sample.
15. The caller recognition apparatus according to claim 14 , wherein to generate the age estimate of the enrollee includes at least one of weighting and offsetting based on the gender.
16. The caller recognition apparatus according to claim 13 , wherein the gender is a gender estimate generated using information collected during the first sample time, the second sample time, or another time.
17. The caller recognition apparatus according to claim 13 , wherein the at least one processor is further configured to:
generate and store a statistical model for the enrollee based on the first sample;
compute the first likelihood score indicating a likelihood that the first sample and the second sample are both from the enrollee,
adjust the first likelihood score based on a calibrated likelihood score determination algorithm to generate the second likelihood score;
determine whether the second likelihood score is greater than a predetermined threshold score; and
if the second likelihood score is greater than the predetermined threshold score, update the statistical model based on the second sample.
18. The device according to claim 16 , further comprising a deep neural network configured to compute the first likelihood score, the deep neural network being trained to discriminate between callers based at least on the statistical model.
19. The caller recognition apparatus according to claim 13 , wherein at least one of the first sample and the second sample is an audio voice sample.
20. The caller recognition apparatus according to claim 13 , wherein at least one of the first sample and the second sample is a facial image sample.
21. The caller recognition apparatus according to claim 13 , wherein the processor is configured to obtain the age of the caller only when the interval is greater than a predetermined time threshold.
22. The caller recognition apparatus according to claim 13 , wherein to generate the age estimate of the caller includes at least one of weighting and offsetting based on the gender.