Graph-based approach for voice authentication

1. A voice biometric authentication system comprising:

a processor and a computer readable medium operably coupled thereto, the computer readable medium comprising a plurality of instructions stored in association therewith that are accessible to, and executable by, the processor, to perform operations which comprise:

(i) receiving a plurality of mono telephonic interactions between customers and agents;

(ii) creating a mapping of the plurality of mono telephonic interactions that illustrates which agent interacted with which customer in each of the interactions;

(iii) determining how many agents each customer interacted with;

(iv) identifying one or more customers an agent has interacted with that have the fewest interactions with other agents based on step (iii);

(v) selecting a predetermined number of interactions of the agent with each of the identified customers; and

(vi) processing the selected predetermined number of interactions of the agent with each of the identified customers into voice prints.

2. The system of claim 1 , wherein processing the selected predetermined number of interactions of the agent with each of the identified customers into voiceprints comprises:

separating each selected interaction into a first speaker component and a second speaker component; and

for each selected interaction, creating a voice print from the first speaker component and a voice print from the second speaker component; and

wherein the operations further comprise:

(vii) comparing the voice print from the first speaker component of a first selected interaction to (1) the voice print from the first speaker component of a second selected interaction and (2) the voice print from the second speaker component of the second selected interaction;

(viii) comparing the voice print from the second speaker component of the first selected interaction to (1) the voice print from the first speaker component of the second selected interaction and (2) the voice print from the second speaker component of the second selected interaction;

(ix) calculating a score indicating similarity between the voice prints in steps (vii) and (viii);

(x) repeating steps (vii), (viii), and (ix) for each successive selected interaction, wherein the comparisons define an analytical path through the selected interactions;

(xi) aggregating the scores for each analytical path; and

(xii) identifying an analytical path having the highest aggregated score as associated with the agent.

3. The system of claim 2 , wherein the operations further comprise creating a voice print of the agent from the voice prints in the identified path.

4. The system of claim 3 , wherein the operations further comprise receiving a mono telephonic interaction between the agent and a customer.

5. The system of claim 4 , wherein the operations further comprise separating the received interaction into a first speaker component and a second speaker component.

6. The system of claim 5 , wherein the operations further comprise comparing the first speaker component and the second speaker component to the voice print of the agent.

7. The system of claim 6 , wherein the operations further comprise:

determining that the first speaker component is associated with the agent;

determining that the second speaker component is associated with the customer; and

enrolling the voice print of the agent in an agent voice print database.

8. The system of claim 7 , wherein the operations further comprise:

creating a customer voice print from the second speaker component; and

enrolling the customer voice print in a customer voice print database.

9. The system of claim 8 , wherein the operations further comprise authenticating a customer based on the enrolled customer voice print.

10. A method, which comprises:

(i) receiving a plurality of mono telephonic interactions between customers and agents;

(ii) creating a mapping of the plurality of mono telephonic interactions that illustrates which agent interacted with which customer in each of the interactions;

(iii) determining how many agents each customer interacted with;

(iv) identifying one or more customers an agent has interacted with that have the fewest interactions with other agents based on step (iii);

(v) selecting a predetermined number of interactions of the agent with each of the identified customers; and

(vi) processing the selected predetermined number of interactions of the agent with each of the identified customers into voice prints.

11. The method of claim 10 wherein processing the selected predetermined number of interactions of the agent with each of the identified customers into voiceprints comprises:

separating each selected interaction into a first speaker component and a second speaker component; and

for each selected interaction, creating a voice print from the first speaker component and a voice print from the second speaker component; and

further comprising:

(vii) comparing the voice print from the first speaker component of a first selected interaction to (1) the voice print from the first speaker component of a second selected interaction and (2) the voice print from the second speaker component of the second selected interaction;

(viii) comparing the voice print from the second speaker component of the first selected interaction to (1) the voice print from the first speaker component of the second selected interaction and (2) the voice print from the second speaker component of the second selected interaction;

(ix) calculating a score indicating similarity between the comparisons in steps (vii) and (viii);

(x) repeating steps (vii), (viii), and (ix) for each successive selected interaction, wherein the comparisons define an analytical path through the selected interactions;

(xi) aggregating the scores for each analytical path; and

(xii) identifying an analytical path having the highest aggregated score as associated with the agent.

12. The method of claim 11 , further comprising creating a voice print of the agent from the voice prints in the identified analytical path.

13. The method of claim 12 , further comprising:

receiving a mono telephonic interaction between the agent and a customer; and

separating the received interaction into a first speaker component and a second speaker component.

14. The method of claim 13 , further comprising comparing the first speaker component and the second speaker component to the voice print of the agent.

15. The method of claim 14 , further comprising:

determining that the first speaker component is associated with the agent;

determining the second speaker component is associated with the customer; and

enrolling the voice print of the agent in an agent voice print database.

16. The method of claim 15 , further comprising:

creating a customer voice print from the second speaker component;

enrolling the customer voice print in a customer voice print database; and

authenticating a customer based on the enrolled customer voice print.

17. A non-transitory computer-readable medium having stored thereon computer-readable instructions executable by a processor to perform operations which comprise:

(i) receiving a plurality of mono telephonic interactions between customers and agents;

(ii) creating a mapping of the plurality of mono telephonic interactions that illustrates which agent interacted with which customer in each of the interactions;

(iii) determining how many agents each customer interacted with;

(iv) identifying one or more customers an agent has interacted with that have the fewest interactions with other agents based on step (iii);

(v) selecting a predetermined number of interactions of the agent with each of the identified customers; and

(vi) processing the selected predetermined number of interactions of the agent with each of the identified customers into voice prints.

18. The non-transitory computer-readable medium of claim 17 , wherein processing the selected predetermined number of interactions of the agent with each of the identified customers into voiceprints comprises:

separating each selected interaction into a first speaker component and a second speaker component; and

for each selected interaction, creating a voice print from the first speaker component and a voice print from the second speaker component; and

wherein the operations further comprise:

(vii) comparing the voice print from the first speaker component of a first selected interaction to (1) the voice print from the first speaker component of a second selected interaction and (2) the voice print from the second speaker component of the second selected interaction;

(viii) comparing the voice print from the second speaker component of the first selected interaction to (1) the voice print from the first speaker component of the second selected interaction and (2) the voice print from the second speaker component of the second selected interaction;

(ix) calculating a score indicating similarity between the comparisons in steps (vii) and (viii);

(x) repeating steps (vii), (viii), and (ix) for each successive selected interaction, wherein the comparisons define an analytical path through the selected interactions;

(xi) aggregating the scores for each path; and

(xii) identifying a path having the highest aggregated score as associated with the agent.

19. The non-transitory computer-readable medium of claim 18 , wherein the operations further comprise creating a voice print of the agent from the voice prints in the identified path.

20. The non-transitory computer-readable medium of claim 19 , wherein the operations further comprise:

receiving a mono telephonic interaction between the agent and a customer;

separating the received interaction into a first speaker component and a second speaker component; and

comparing the first speaker component and the second speaker component to the voice print of the agent.

21. The non-transitory computer-readable medium of claim 20 , wherein the operations further comprise:

determining that the first speaker component is associated with the agent;

determining the second speaker component is associated with the customer; and

enrolling the voice print of the agent in an agent voice print database.

22. The non-transitory computer-readable medium of claim 21 , wherein the operations further comprise:

creating a customer voice print from the second speaker component;

enrolling the customer voice print in a customer voice print database; and

authenticating a customer based on the enrolled customer voice print.