Enhanced diarization systems, media and methods of use

• US 9,571,652 B1
• Filed: 04/09/2012
• Issued: 02/14/2017
• Est. Priority Date: 04/21/2005
• Status: Active Grant

• Current Assignee: Verint Americas Incorporated (Verint Systems Incorporated)
• Original Assignee: Verint Americas Incorporated (Verint Systems Incorporated)
• Inventors: Efron, Andrew S., Zeppenfeld, Torsten, Gutierrez, Richard, Guerra, Lisa, Rajakumar, Anthony
• Primary Examiner(s): Elahee, Md S
• Application Number:
  US13/442,767
• Time in Patent Office: 1,772 Days
• Field of Search: 379 8801- 8804, 379/114.14, 37926501-2661, 381/17, 704/260, 704/272, 704/246, 704/273, 705/38, 370/521
• US Class Current: 1/1
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques
• CPC Class Codes: H04L 9/3231 Biological data, e.g. fingerprint, voice or retina network architectures or network communication protocols for supporting authentication of entities using biometrical features in a packet data network H04L63/0861
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 2203/6027 Fraud preventions
• CPC Class Codes: H04M 3/2281 Call monitoring, e.g. for law enforcement purposes; Call tracing; Detection or prevention of malicious calls
• CPC Class Codes: H04M 3/51 Centralised call answering arrangements requiring operator intervention , e.g. call or contact centers for telemarketing
• CPC Class Codes: H04M 3/523 with call distribution or queueing

Abstract

Enhanced diarization systems and methods of use are provided herein. Some exemplary methods may include applying one or more rules that affect separation of the call audio data into segments, the rules being associated with the at least one call schema, separating the call audio data into segments according to the one or more rules, grouping segments of call audio data associated with a speaker, and storing in a storage media an identifier and the grouped segments for the speaker.

Claims

1. A method for processing call audio data, the method comprising:

receiving a stream of call audio data collected during a call event between an agent and a caller;

separating the stream of call audio data into segments based on a call schema;

extracting, for each segment, a channel model that describes how the call audio data has been affected by the channel through which the call audio data has been communicated, wherein an agent'"'"'s channel model is distinct from a caller'"'"'s channel model;

comparing channel models for each segment;

grouping segments corresponding to the caller based on the comparison;

identifying the caller as a fraudster by comparing the channel models in the group of segments corresponding to the caller to channel models stored in a fraudster database;

transmitting to the agent a notification if the caller is a fraudster; and

enrolling the grouped segments into the fraudster database.

2. The method of claim 1, wherein the channel model represents a communication profile of a voice changer.

3. The method of claim 1, wherein the channel model indicates that the call audio data originated from a cellular telephone.

4. A method for diarizing call audio data, the method comprising:

receiving a call schema for call audio data; and

diarizing the call audio data by;

applying one or more rules that affect separation of the call audio data into segments, the rules being associated with the received call schema,

separating the call audio data into segments according to the one or more rules,

extracting a channel model from each segment that describes how the call audio data has been affected by the channel through which the call audio data has been communicated,

grouping the segments by speaker based on each speaker'"'"'s channel model,

identifying a speaker as a fraudster by comparing the channel models in the group of segments corresponding to the speaker to channel models stored in a fraudster database, and

creating an audio file for each group of segments.

5. The method of claim 4, wherein the channel model indicates that the call audio data originated from a cellular telephone.

6. A system for diarizing call audio data, the system comprising:

a memory for storing an executable instructions that diarize call audio data;

a processor that executes the instructions;

a diarization module stored in memory and executable by the processor to;

receive a call schema for call audio data,

apply one or more rules that affect separation of the call audio data into segments, the rules being associated with the received call schema,

separate the call audio data into segments according to the one or more rules,

group the segments by speaker based on each speaker'"'"'s channel model, and

create an audio file for each group of segments; and

an analysis module stored in memory and executable by the processor to;

determine a channel model for each segment, wherein the channel model describes how the call audio data has been affected by the channel through which the call audio data has been communicated, and

compare the channel models in a group of segments corresponding to a speaker to channel models stored in a fraudster database to identify the speaker as a fraudster.

7. The system according to claim 6, further comprising a database module stored in memory and executable by the processor to store in a non-transitory storage media the grouped segments for each speaker.

8. The system according to claim 6, further comprising an enrollment module stored in memory and executable by the processor to store the speaker'"'"'s channel model in a fraudster database if the speaker is associated with an instance of fraud.

9. The method of claim 6, wherein the channel model indicates that the call audio data originated from a cellular telephone.

10. A non-transitory computer readable storage media having a program embodied thereon, the program being executable by a processor to perform a method for diarizing call audio data, the method comprising:

determining at least one call schema for call audio data; and

diarizing the call audio data by;

applying one or more rules that affect separation of the call audio data into segments, the rules being associated with the at least one call schema; and

separating the call audio data into segments according to the one or more rules;

extracting a channel model from each segment that describes how the call audio data has been affected by the channel through which the call audio data has been communicated;

grouping the segments by speaker based on each speaker'"'"'s channel model;

identifying a speaker as a fraudster by comparing the channel models in the group of segments corresponding to the speaker to channel models stored in a fraudster database; and

creating an audio file for each group of segments.

11. The method of claim 10, wherein the channel model indicates that the call audio data originated from a cellular telephone.