ENROLLMENT AND AUTHENTICATION OVER A PHONE CALL IN CALL CENTERS
Embodiments described herein provide for a voice biometrics system execute machine-learning architectures capable of passive, active, continuous, or static operations, or a combination thereof. Systems passively and/or continuously, in some cases in addition to actively and/or statically, enrolling speakers. The system may dynamically generate and update profiles corresponding to end-users who contact a call center. The system may determine a level of enrollment for the enrollee profiles that limits the types of functions that the user may access. The system may update the profiles as new contact events are received or based on certain temporal triggering conditions.
1 . A computer-implemented method comprising:
receiving, by a computer, one or more enrollment inputs for an enrolled speaker, the one or more enrollment inputs comprising an enrollment audio signal and enrollment data;
generating, by the computer, an enrolled voiceprint for the enrolled speaker using the one or more enrollment signals and an enrollment level for the enrolled speaker based upon the enrollment data;
receiving, by the computer, one or more inbound speaker inputs for an inbound speaker, the one or more inbound speaker inputs comprising an inbound audio signal and inbound contact data; and
generating, by the computer, an authentication score for the inbound speaker using the enrolled voiceprint, the enrollment level, and an inbound voiceprint for the inbound speaker based on the inbound audio signal.
2 . The method according to claim 1 , wherein an input includes at least one of an inbound input and an enrollment input, and wherein the input includes one or more knowledge responses.
3 . The method according to claim 1 , further comprising transmitting, by the computer, a prompt for an input to an end-user device, wherein the input includes at least one of an inbound input and an enrollment input, and wherein the input includes a response to the prompt.
4 . The method according to claim 1 , further comprising:
extracting, by the computer, an enrollment deviceprint using the enrollment data;
extracting, by the computer, an inbound deviceprint using the inbound contact data; and
generating, by the computer, a device similarity score using the inbound deviceprint and the enrollment deviceprint, wherein the authentication score for the inbound speaker is further based upon the the device similarity score.
5 . The method according to claim 1 , further comprising:
extracting, by the computer, an enrollment behaviorprint using the enrollment data;
extracting, by the computer, an inbound behaviorprint using the inbound contact data; and
generating, by the computer, a behavior similarity score using the inbound behaviorprint and the enrollment behaviorprint, wherein the authentication score for the inbound speaker is further based upon the behavior similarity score.
6 . The method according to claim 1 , further comprising authenticating, by the computer, the inbound speaker as the enrolled speaker based in part upon determining that the authentication score satisfies an authentication threshold score.
7 . The method according to claim 1 , further comprising authenticating, by the computer, the inbound speaker as the enrolled speaker based in part upon determining that the authentication level satisfies the enrollment level.
8 . The method according to claim 1 , further comprising generating, by the computer, a risk score for the enrollment input based upon at least one of a global risk factor and a local risk factor, wherein the enrollment level is based in part upon the risk score.
9 . The method according to claim 1 , wherein the enrollment inputs includes a plurality of types of data, and wherein the enrollment level is determined based upon a relative weight associated a corresponding type of data of the plurality of types of data.
10 . The method according to claim 9 , wherein the plurality of types of data of the enrollment inputs include at least one of: a weak knowledge based authentication value, a strong knowledge based authentication value, a one-time password, a push notification response, and an embedding vector.
11 . A system comprising:
a database configured to store a plurality of enrollment inputs; and
a computer comprising a processor configured to:
receive one or more enrollment inputs for an enrolled speaker, the one or more enrollment inputs comprising an enrollment audio signal and enrollment data;
generate an enrolled voiceprint for the enrolled speaker using the one or more enrollment signals and an enrollment level for the enrolled speaker based upon the enrollment data;
receive one or more inbound speaker inputs for an inbound speaker, the one or more inbound speaker inputs comprising an inbound audio signal and inbound contact data; and
generate an authentication score for the inbound speaker using the enrolled voiceprint, the enrollment level, and an inbound voiceprint for the inbound speaker based on the inbound audio signal.
12 . The system according to claim 11 , wherein an enrollment input includes at least one of an inbound input and an enrollment input, and wherein the input includes one or more knowledge responses.
13 . The system according to claim 1 , wherein the computer is further configured to transmit a prompt for an input to an end-user device, and wherein the input includes at least one of an inbound input and an enrollment input, and wherein the input includes a response to the prompt.
14 . The system according to claim 1 , wherein the computer is further configured to:
extract an enrollment deviceprint using the enrollment data;
extract an inbound deviceprint using the inbound contact data; and
generate a device similarity score using the inbound deviceprint and the enrollment deviceprint, wherein the authentication score for the inbound speaker is further based upon the the device similarity score.
15 . The system according to claim 1 , wherein the computer is further configured to:
extract an enrollment behaviorprint using the enrollment data;
extract an inbound behaviorprint using the inbound contact data; and
generate a behavior similarity score using the inbound behaviorprint and the enrollment behaviorprint, wherein the authentication score for the inbound speaker is further based upon the behavior similarity score.
16 . The system according to claim 1 , wherein the computer is further configured to authenticate the inbound speaker as the enrolled speaker based in part upon determining that the authentication score satisfies an authentication threshold score.
17 . The system according to claim 1 , wherein the computer is further configured to authenticate the inbound speaker as the enrolled speaker based in part upon determining that the authentication level satisfies the enrollment level.
18 . The system according to claim 1 , wherein the computer is further configured to generate a risk score for the enrollment input based upon at least one of a global risk factor and a local risk factor, wherein the enrollment level is based in part upon the risk score.
19 . The system according to claim 11 , wherein the enrollment inputs includes a plurality of types of data, and wherein the enrollment level is determined based upon a relative weight associated a corresponding type of data of the plurality of types of data.
20 . The system according to claim 19 , wherein the plurality of types of data of the enrollment inputs include at least one of: a weak knowledge based authentication value, a strong knowledge based authentication value, a one-time password, a push notification response, and an embedding vector.