Method and system for screening using voice data and metadata

• US 8,311,826 B2
• Filed: 10/20/2011
• Issued: 11/13/2012
• Est. Priority Date: 04/21/2005
• Status: Active Grant

• Current Assignee: Verint Americas Incorporated (Verint Systems Incorporated)
• Original Assignee: Victria Incorporated (Verint Systems Incorporated)
• Inventors: Rajakumar, Anthony
• Primary Examiner(s): Albertalli, Brian
• Application Number:
  US13/278,067
• Publication Number:
  US 20120054202A1
• Time in Patent Office: 390 Days
• Field of Search: None
• US Class Current: 704/246
• CPC Class Codes: G06Q 20/24 Credit schemes, i.e. "pay after"
• CPC Class Codes: G06Q 20/4014 Identity check for transactions
• CPC Class Codes: G06Q 30/0185 Product, service or business identity fraud
• CPC Class Codes: G06Q 40/03 Credit; Loans; Processing thereof
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques

Abstract

In one embodiment, the invention provides a method, comprising: (a) maintaining a database of known fraudsters, wherein each fraudster has a voice signature and metadata associated therewith; (b) performing a screening operation to match a candidate with a fraudster in the database based on matching a voice signature and metadata for the candidate with a voice signature and metadata for the fraudster in the database.

Claims

1. A system comprising:

a processor; and

a memory coupled to the processor, the memory storing instructions which when executed by the processor causes the system to perform a method comprising;

maintaining a database of fraudster records for known fraudsters, each fraudster record including a voice signature and metadata associated therewith, the database of fraudster records comprising a plurality of segmented databases according to the metadata;

removing an inactive fraudster record from a segmented database after a predefined period; and

performing a screening operation to match a candidate and a known fraudster based on matching a voice signature and metadata for the candidate with a voice signature and metadata in the a fraudster record in the database.

2. The system of claim 1, wherein segmentation of the segmented databases is based on industry type.

3. The system of claim 1, wherein segmented databases are sorted based on frequency of matching of fraudster records in each segmented databases.

4. The system of claim 3 wherein a segmented database having frequently detected voiceprints is at the head of the sorted segmented databases.

5. The system of claim 1, further comprising extracting metadata for the candidate using speech recognition.

6. The system of claim 1, wherein a probability of matching the voice signature and metadata for the candidate is based on a distance of a location in the metadata for the candidate from a location in the metadata in the fraudster record.

7. A method, comprising:

collecting a voice sample and related metadata without regard to association of the voice sample and related metadata with fraud;

creating a database of known fraudsters based on an association of metadata with fraud using an enrollment engine, the database including the collected voice sample and related metadata;

removing the collected voice sample and related metadata from the database of known fraudsters after the collected voice sample or related metadata have not been matched during a period of time;

receiving a candidate voice signature and metadata;

comparing the received candidate voice signature and metadata for a match with the collected voice sample and related metadata using a screening engine; and

assigning a score based on the comparison.

8. The method of claim 7, further comprising segmenting the database into a plurality of segmented databases according to location information in the metadata.

9. The method of claim 8, wherein the screening engine performs the comparison on one of the segmented databases.

10. The method of claim 7, wherein the database is sorted based on a number of matches.

11. The method of claim 7, wherein metadata is extracted using speech recognition.

12. The method of claim 7, further comprising:

segmenting the database of known fraudsters into a plurality of segmented databases;

determining if the collected voice sample and related metadata is inactive in a first segmented database; and

searching the remaining databases to determine if the collected voice sample and related metadata is active in a second database.

13. A system for detecting fraudsters comprising:

an audio capture device configured to collect voice samples;

a database of known fraudster entries, each fraudster entry including a fraudster voice signature and metadata;

an enrollment engine configured to remove an inactive fraudster entry from a first segmented database and search the remaining segmented databases to determine if the removed fraudster entry is active in a second segmented database; and

a screening engine configured to search the database for a match of a candidate voice sample and metadata with a matching fraudster voice signature and metadata, and to assign a screening score based on the match.

14. The system of claim 13, further comprising an enrollment engine configured to build the fraudster voice signature from a voice sample and related metadata collected without regard to fraud, the voice sample captured using the audio capture device.

15. The system of claim 13, further wherein screening engine is further configured to initiate a telephone call to the candidate and to collect the candidate voice sample using the audio capture device from responses to questions posed to the candidate during the telephone call.

16. The system of claim 13, further comprising another screening engine configured to share the database of known fraudster entries.

17. The system of claim 13, wherein the database includes a sorted plurality of segmented databases, and the screening engine is further configured to search a segmented database according to the order of the segmented databases.

18. A method, comprising:

collecting a voice sample and related metadata without regard to association of the voice sample and related metadata with fraud;

creating a database of known fraudsters based on an association of metadata with fraud using an enrollment engine, the database including the collected voice sample and related metadata;

segmenting the database of known fraudsters into a plurality of segmented databases;

determining if the collected voice sample and related metadata is inactive in a first segmented database; and

searching the remaining databases to determine if the collected voice sample and related metadata is active in a second database;

receiving a candidate voice signature and metadata;

comparing the received candidate voice signature and metadata for a match with the collected voice sample and related metadata using a screening engine; and

assigning a score based on the comparison.