Systems and methods for cluster-based voice verification

• US 10,003,688 B1
• Filed: 02/08/2018
• Issued: 06/19/2018
• Est. Priority Date: 02/08/2018
• Status: Active Grant

• Current Assignee: Capital One Services LLC (Capital One Financial Corporation)
• Original Assignee: Capital One Services LLC (Capital One Financial Corporation)
• Inventors: Walters, Austin, Goodsitt, Jeremy, Abdi Taghi Abad, Fardin
• Primary Examiner(s): Milord, Marceau
• Application Number:
  US15/891,712
• Time in Patent Office: 131 Days
• Field of Search: 455415, 4554141, 4554144
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
• CPC Class Codes: G10L 25/06 the extracted parameters being correlation coefficients
• CPC Class Codes: G10L 25/27 characterised by the analysis technique
• CPC Class Codes: G10L 25/30 using neural networks
• CPC Class Codes: G10L 25/51 for comparison or discrimination
• CPC Class Codes: H04M 2201/41 using speaker recognition speaker recognition per se G10L17/00
• CPC Class Codes: H04M 2203/556 Statistical analysis and interpretation
• CPC Class Codes: H04M 2203/6045 Identity confirmation
• CPC Class Codes: H04M 3/42042 Notifying the called party of information on the calling party details within substation equipment H04M1/57, signalling details H04Q3/72

Abstract

Systems for caller identification and authentication may include an authentication server. The authentication server may be configured to receive audio data including speech of a plurality of telephone calls, use audio data for at least a subset of the plurality of telephone calls to populate a plurality of word clusters each associated with a specific demographic, and/or use audio data for at least one of the plurality of telephone calls to identify the telephone caller making the telephone call based on determining a most similar word cluster of the plurality of word clusters to the audio data of the caller.

Claims

1. A method of authenticating a telephone caller, the method comprising:

receiving, by a processor of an authentication server, audio data including speech of the telephone caller;

analyzing, by the processor, the audio data to identify a plurality of words from the speech of the telephone caller and to identify an occurrence frequency for each of the plurality of words;

comparing, by the processor, the plurality of words and the occurrence frequencies to a plurality of word clusters, each word cluster comprising a plurality of associated words and an occurrence frequency for each of the plurality of associated words, and each word cluster being associated with one of a plurality of demographics;

determining, by the processor, a most similar word cluster of the plurality of word clusters to the audio data based on a similarity of the plurality of words and the plurality of associated words of the most similar cluster and a similarity of the occurrence frequencies of the plurality of words and the occurrence frequencies of the plurality of associated words of the most similar cluster;

receiving, by the processor, a purported identity of the telephone caller, the purported identity including caller demographic data;

comparing, by the processor, the caller demographic data to the demographic associated with the most similar word cluster; and

identifying, by the processor, the telephone caller as at least one of;

likely having the purported identity in response to determining the caller demographic data matches the demographic associated with the most similar word cluster, and

unlikely to have the purported identity in response to determining the caller demographic data matches a demographic associated with a word cluster different from the most similar word cluster.

2. The method of claim 1, further comprising:

analyzing, by the processor, the audio data to identify at least one acoustic characteristic of the speech of the telephone caller; and

comparing, by the processor, the at least one acoustic characteristic of the speech of the telephone caller to the plurality of word clusters, each word cluster further comprising at least one associated acoustic characteristic;

wherein the determining, by the processor, the most similar word cluster of the plurality of word clusters to the audio data is further based on a similarity of the at least one acoustic characteristic of the speech of the telephone caller and the at least one associated acoustic characteristic of the most similar cluster.

3. The method of claim 2, wherein the analyzing, by the processor, the audio data to identify at least one acoustic characteristic of the speech of the telephone caller comprises:

correlating, by the processor, each of a plurality of portions of an acoustic or frequency component of the audio data with each of at least a subset of the plurality of words; and

determining, by the processor, at least one acoustic characteristic for how the telephone caller says at least one of the subset of the plurality of words based on the portion of the acoustic or frequency component of the audio data correlated with the at least one of the subset of the plurality of words.

4. The method of claim 1, wherein:

the caller demographic data comprises current caller demographic data and historical caller demographic data;

determining the caller demographic data matches the demographic associated with the most similar word cluster comprises determining at least one of the current caller demographic data and the historical caller demographic data matches the demographic associated with the most similar word cluster; and

determining the caller demographic data matches the demographic associated with the word cluster different from the most similar word cluster comprises determining at least one of the current caller demographic data and the historical caller demographic data matches the demographic associated with the word cluster different from the most similar word cluster.

5. The method of claim 1, further comprising:

receiving, by the processor, a threat score for the telephone caller;

wherein the identifying, by the processor, the telephone caller as likely having the purported identity comprises lowering the threat score or maintaining the threat score as received.

6. The method of claim 1, further comprising:

receiving, by the processor, a threat score for the telephone caller;

wherein the identifying, by the processor, the telephone caller as unlikely to have the purported identity comprises raising the threat score.

7. A method of identifying a telephone caller, the method comprising:

receiving, by a processor of an authentication server, audio data including speech of a plurality of telephone calls;

for at least a subset of the plurality of telephone calls, determining, by the processor, demographic data for a telephone caller making the telephone call;

for at least the subset of the plurality of telephone calls, analyzing, by the processor, the audio data to identify a plurality of words from the speech of the telephone caller;

receiving, by the processor, a plurality of word clusters, each word cluster associated with a specific demographic;

populating, by the processor, at least one word cluster with at least a subset of the plurality of words from the speech of each telephone caller associated with the specific demographic based on the demographic data for the telephone caller;

for each word cluster, determining, by the processor, a plurality of associated words from among at least the subset of the plurality of words and an occurrence frequency for each of the plurality of associated words; and

for at least one of the plurality of telephone calls;

analyzing, by the processor, the audio data to identify a plurality of words from the speech of the telephone caller and to identify an occurrence frequency for each of the plurality of words,

comparing, by the processor, the plurality of words from the speech of the telephone caller and the occurrence frequency for each of the plurality of words from the speech of the telephone caller to the plurality of word clusters,

based on the comparing, identifying, by the processor, a most similar word cluster of the plurality of word clusters to the audio data based on a similarity of the plurality of words from the speech of the telephone caller and the plurality of associated words of the most similar cluster and a similarity of the occurrence frequencies of the plurality of words from the speech of the telephone caller and the occurrence frequencies of the plurality of associated words of the most similar cluster, and

determining, by the processor, a caller demographic of the telephone caller, the caller demographic being the same as the demographic of the most similar word cluster.

8. The method of claim 7, further comprising, for at least the subset of the plurality of telephone calls:

analyzing, by the processor, the audio data to identify at least one acoustic characteristic of the speech of the telephone caller; and

populating, by the processor, at least one word cluster with at least a subset of the at least one acoustic characteristic of the speech of each telephone caller associated with the specific demographic based on the demographic data for the telephone caller.

9. The method of claim 8, wherein the analyzing, by the processor, the audio data to identify at least one acoustic characteristic of the speech of the telephone caller comprises:

correlating, by the processor, each of a plurality of portions of an acoustic or frequency component of the audio data with each of at least a subset of the plurality of words; and

determining, by the processor, at least one acoustic characteristic for how the telephone caller says at least one of the subset of the plurality of words based on the portion of the acoustic or frequency component of the audio data correlated with the at least one of the subset of the plurality of words.

10. The method of claim 8, further comprising, for the at least one of the plurality of telephone calls:

analyzing, by the processor, the audio data to identify at least one acoustic characteristic of the speech of the telephone caller;

comparing, by the processor, the at least one acoustic characteristic of the speech of the telephone caller to the plurality of word clusters;

wherein the determining, by the processor, the most similar word cluster of the plurality of word clusters to the audio data is further based on a similarity of the at least one acoustic characteristic of the speech of the telephone caller and the at least one associated acoustic characteristic of the most similar cluster.

11. The method of claim 10, wherein the analyzing, by the processor, the audio data to identify at least one acoustic characteristic of the speech of the telephone caller comprises:

correlating, by the processor, each of a plurality of portions of an acoustic or frequency component of the audio data with each of at least a subset of the plurality of words; and

determining, by the processor, at least one acoustic characteristic for how the telephone caller says at least one of the subset of the plurality of words based on the portion of the acoustic or frequency component of the audio data correlated with the at least one of the subset of the plurality of words.

12. The method of claim 7, further comprising:

receiving, by the processor, a purported identity of the telephone caller, the purported identity including a purported demographic;

comparing, by the processor, the caller demographic to the purported demographic; and

identifying, by the processor, the telephone caller as at least one of;

likely having the purported identity in response to determining the caller demographic matches the purported demographic, and

unlikely to have the purported identity in response to determining the caller demographic matches a demographic other than the purported demographic.

13. The method of claim 12, wherein:

the purported identity comprises current caller demographic data and historical caller demographic data;

determining the caller demographic matches the purported demographic comprises determining at least one of the current caller demographic data and the historical caller demographic data matches the caller demographic; and

determining the caller demographic data matches the demographic other than the purported demographic comprises determining neither of the current caller demographic data and the historical caller demographic data matches the caller demographic.

14. The method of claim 12, further comprising:

receiving, by the processor, a threat score for the telephone caller;

wherein the identifying, by the processor, the telephone caller as likely having the purported identity comprises lowering the threat score or maintaining the threat score as received.

15. The method of claim 12, further comprising:

receiving, by the processor, a threat score for the telephone caller;

wherein the identifying, by the processor, the telephone caller as unlikely to have the purported identity comprises raising the threat score.

16. A system for caller identification and authentication, the system comprising:

a telephony recorder configured to record audio data for calls placed to at least one phone number;

an authentication server comprising a processor and a non-transitory memory, the memory storing instructions that, when executed by the processor, cause the processor to perform processing comprising;

receiving audio data including speech of a plurality of telephone calls;

using audio data for at least a subset of the plurality of telephone calls to populate a plurality of word clusters, each word cluster being associated with a specific demographic, the populating of the plurality of word clusters comprising;

for each of the subset of the plurality of telephone calls, determining demographic data for a telephone caller making the telephone call, and analyzing the audio data to identify a plurality of words from the speech of the telephone caller, and

populating at least one word cluster with at least a subset of the plurality of words from the speech of each telephone caller associated with the specific demographic based on the demographic data for the telephone caller; and

using audio data for at least one of the plurality of telephone calls to identify the telephone caller making the telephone call, the identifying comprising;

analyzing the audio data to identify a plurality of words from the speech of the telephone caller and to identify an occurrence frequency for each of the plurality of words,

comparing, the plurality of words and the occurrence frequencies to the plurality of word clusters,

determining a most similar word cluster of the plurality of word clusters to the audio data based on a similarity of the plurality of words and the plurality of associated words of the most similar cluster and a similarity of the occurrence frequencies of the plurality of words and occurrence frequencies of the plurality of associated words of the most similar cluster,

receiving a purported identity of the telephone caller, the purported identity including caller demographic data,

determining whether the caller demographic data matches the demographic associated with the most similar word cluster, and

identifying the telephone caller as;

likely having the purported identity in response to determining that the caller demographic data matches the demographic associated with the most similar word cluster, or

unlikely to have the purported identity in response to determining that the caller demographic data does not match the demographic associated with the most similar word cluster.

17. The system of claim 16, wherein the instructions further cause the processor to perform processing comprising, for at least the subset of the plurality of telephone calls:

analyzing the audio data to identify at least one acoustic characteristic of the speech of the telephone caller; and

populating at least one word cluster with at least a subset of the at least one acoustic characteristic of the speech of each telephone caller associated with the specific demographic based on the demographic data for the telephone caller.

18. The system of claim 17, wherein the analyzing of the audio data to identify at least one acoustic characteristic of the speech of the telephone caller comprises:

correlating each of a plurality of portions of an acoustic or frequency component of the audio data with each of at least a subset of the plurality of words; and

determining at least one acoustic characteristic for how the telephone caller says at least one of the subset of the plurality of words based on the portion of the acoustic or frequency component of the audio data correlated with the at least one of the subset of the plurality of words.

19. The system of claim 16, wherein the instructions further cause the processor to perform processing comprising, for the at least one of the plurality of telephone calls:

analyzing the audio data to identify at least one acoustic characteristic of the speech of the telephone caller;

comparing the at least one acoustic characteristic of the speech of the telephone caller to the plurality of word clusters;

wherein the determining the most similar word cluster of the plurality of word clusters to the audio data is further based on a similarity of the at least one acoustic characteristic of the speech of the telephone caller and the at least one associated acoustic characteristic of the most similar cluster.

20. The system of claim 19, wherein the analyzing the audio data to identify at least one acoustic characteristic of the speech of the telephone caller comprises:

correlating each of a plurality of portions of an acoustic or frequency component of the audio data with each of at least a subset of the plurality of words; and

determining at least one acoustic characteristic for how the telephone caller says at least one of the subset of the plurality of words based on the portion of the acoustic or frequency component of the audio data correlated with the at least one of the subset of the plurality of words.