Biometrics

Use the biometrics node as a speaker identification system.

Task Type	Identification
Voice	The caller speaks now to identify themselves. If the caller has already identified themselves and that data is captured in an audio variable, use the audio variable to identify the caller.
Workflow	Use an audio variable to identify the caller. This requires that the caller has already identified themselves and that data is captured in an audio variable.

Task Type

Identification

Voice

The caller speaks now to identify themselves.

If the caller has already identified themselves and that data is captured in an audio variable, use the audio variable to identify the caller.

Workflow

Use an audio variable to identify the caller. This requires that the caller has already identified themselves and that data is captured in an audio variable.

The biometrics node has three actions.

Action	Description
Enroll	Caller records a voiceprint. This enrolls the caller in the system.
Verify	Caller's voice is compared to the voiceprint.
Delete	Delete the voiceprint.

Enroll

The caller records a voiceprint. This enrolls the caller in the system.

From the Name field, give the node a name.
From the Action field, select Enroll.

Change the Save Recorded Utterances setting if desired.

Setting	Description
	Save the utterance as an audio file on the Studio platform. Convert the audio file to a voiceprint. Save the voiceprint. Save the utterance during the early stages of implementation to assist with tuning and resolving errors. Save the utterance to: Troubleshoot issues such as false accepts and false rejects. Provide auditable evidence. Allow seamless upgrade to new technology. Provide alternative data for calibration and tuning. Provide an alternative way to re-enroll after tuning. Provide data for calibration and tuning. Provide data for re-enrolling (and active learning) after tuning.
	Do not save the utterance. Save the voiceprint.

Setting

Description

Save the utterance as an audio file on the Studio platform. Convert the audio file to a voiceprint. Save the voiceprint.

Save the utterance during the early stages of implementation to assist with tuning and resolving errors.

Save the utterance to:

Troubleshoot issues such as false accepts and false rejects.
Provide auditable evidence.
Allow seamless upgrade to new technology.
Provide alternative data for calibration and tuning.
Provide an alternative way to re-enroll after tuning.
Provide data for calibration and tuning.
Provide data for re-enrolling (and active learning) after tuning.

Do not save the utterance. Save the voiceprint.

Change the Voiceprint Type, if desired.

Digit is initially selected and works well with voiceprints. Refer to the biometrics documentation for more details on voiceprint types.

Voiceprint Type	Description
Text Independent	For enrollment, the voiceprint should contain at least ten seconds of net conversational speech. For verification, the voiceprint should contain at least four seconds of net conversational speech.
Digit	The voiceprint should be a sequence of at least seven digits, and preferably eight to ten. For enrollment, the voiceprint should contain at least three utterances. For verification, the voiceprint should specify the same sequence of digits as at enrollment. For a voiceprint, you might choose to use an account number, a telephone number, or a standard sequence as in counting.
Phrase	The voiceprint should be a phrase of at least four words, and preferably more. For enrollment, the voiceprint should contain at least three utterances. For verification, the voiceprint should specify the same word sequence as at enrollment. For a voiceprint, you might choose to use your postal address, or a memorable phrase, such as, "my voice is my password".
Date	For enrollment, the voiceprint should contain at least three utterances. For verification, the voiceprint should specify the same date as at enrollment. For a voiceprint, you might choose to use your date of birth.

Repeat the following for the First Utterance, Second Utterance, and Third Utterance.

Setting

Description

Use Recorded Audio

(Voice tasks only)

Prompt the caller to speak now to identify themselves.

In the Caller Instructions field, specify the prompt to play to the caller.

Guide the caller, taking into consideration the voiceprint type. See step 4.

The intention is to capture three samples of the caller's voice (utterances) as a voiceprint.

Ask the caller to repeat the utterance at each prompt.

The utterance should be easy for the caller to remember. When the caller verifies their identity (Verify), the caller repeats the utterance, or part of the utterance, dependent on the voiceprint type.

Utterance	Example prompts where the voiceprint type is digit
First Utterance	After the beep, speak a sequence of ten digits. Ensure that this is an easy sequence for you to remember, so you can repeat the sequence for future identify verification. When you have finished, press the pound key.
Second Utterance	Repeat the sequence, followed by the pound key.
Third Utterance	Repeat the sequence one last time, followed by the pound key.

Change the Record Duration, as desired.

The record duration establishes the maximum recording time (in seconds) the caller has to record the voiceprint. The default is the maximum limit.

Use Audio Variable

The caller has already identified themselves on the call and that data is captured in a variable. Typically, the Cloud STT node or record node generates the recording.

Select the variable storing the recording of the voice enrollment. Enrollment requirements depend on the voiceprint type. See step 4.

Use an audio variable to avoid repeatedly asking for the same information. For example, if enrollment is by account number and a variable captured the caller speaking their account number earlier in the call, use that variable to authenticate the caller.

Open the Event Handler tab.

Field	Description
No Input Event Handler	Select a task canvas. The call routes to the canvas if no input is detected from the caller after multiple attempts. The Count value tells Studio how many attempts to run before triggering the event handler. If not using an existing canvas, enter a name for a new canvas.
No Match Event Handler	Select a task canvas. The call routes to the canvas if no match is detected from the caller after multiple attempts. The Count value tells Studio how many attempts to run before triggering the event handler. If not using an existing canvas, enter a name for a new canvas.

Open the Results tab.
1. Select the variable to store the Enroll ID. The biometrics actions (verify and delete) use the enroll ID variable.
2. Select the variable to store the Enroll Result.
3. Select the variable to store the Enroll Message.
If the variables have not been created yet, type their names in the associated fields.

Verify

Compare the caller's voice to the voiceprint.

From the Name field, give the node a name.
From the Action field, select Verify.

Change the Save Recorded Utterances setting if desired.

Setting	Description
	Save the utterance as an audio file on the Studio platform. Convert the audio file to a voiceprint. Save the utterance during the early stages of implementation to assist with tuning and resolving errors. Save the utterance to: Troubleshoot issues such as false accepts and false rejects. Provide auditable evidence. Allow seamless upgrade to new technology. Provide alternative data for calibration and tuning. Provide an alternative way to re-enroll after tuning. Provide data for calibration and tuning. Provide data for re-enrolling (and active learning) after tuning.
	Do not save the utterance.

Setting

Description

Save the utterance as an audio file on the Studio platform. Convert the audio file to a voiceprint.

Save the utterance during the early stages of implementation to assist with tuning and resolving errors.

Save the utterance to:

Troubleshoot issues such as false accepts and false rejects.
Provide auditable evidence.
Allow seamless upgrade to new technology.
Provide alternative data for calibration and tuning.
Provide an alternative way to re-enroll after tuning.
Provide data for calibration and tuning.
Provide data for re-enrolling (and active learning) after tuning.

Do not save the utterance.

Select the Voiceprint Type specified at enrollment.

Voiceprint Type	Description
Text Independent	For verification, the voiceprint should contain at least four seconds of net conversational speech.
Digit	The voiceprint should be a sequence of at least seven digits, and preferably eight to ten. For verification, the voiceprint should specify the same sequence of digits as at enrollment. Example voiceprints that suit a voiceprint, you might choose to use an account number, a telephone number, or a standard sequence as in counting.
Phrase	The voiceprint should be a phrase of at least four words, and preferably more. For verification, the voiceprint should specify the same word sequence as at enrollment. For a voiceprint, you might choose to use your postal address, or a memorable phrase, such as, "my voice is my password".
Date	For verification, the voiceprint should specify the same date as at enrollment. For a voiceprint, you might choose to use your date of birth.

Change the Verification Utterance setting if desired.

Setting	Description
Use Recorded Audio (Voice tasks only)	Prompt the caller to speak now to identify themselves. In the Caller Instructions field, specify the prompt to play to the caller. Guide the caller, taking into consideration the voiceprint type. See step 4. Example: If the voiceprint type is digit, this is an example prompt: After the beep, repeat the same sequence of digits you spoke when you enrolled in the system, followed by the pound key. Change the Record Duration, as desired. The record duration establishes the maximum recording time (in seconds) the caller has to verify the voiceprint. The default is the maximum limit.
Use Audio Variable	The caller has already identified themselves on the call and that data is captured in a variable. Typically, the Cloud STT node or record node generates the recording. Select the variable storing the recording of the voice verification. Verification requirements depend on the voiceprint type. See step 4. Use an audio variable to avoid repeatedly asking for the same information. For example, if identification is by account number and a variable captured the caller speaking their account number earlier in the call, use that variable to authenticate the caller.

Setting

Description

Use Recorded Audio

(Voice tasks only)

Prompt the caller to speak now to identify themselves.

In the Caller Instructions field, specify the prompt to play to the caller.

Guide the caller, taking into consideration the voiceprint type. See step 4.

Example:
If the voiceprint type is digit, this is an example prompt:

After the beep, repeat the same sequence of digits you spoke when you enrolled in the system, followed by the pound key.
Change the Record Duration, as desired.

The record duration establishes the maximum recording time (in seconds) the caller has to verify the voiceprint. The default is the maximum limit.

Use Audio Variable

The caller has already identified themselves on the call and that data is captured in a variable. Typically, the Cloud STT node or record node generates the recording.

Select the variable storing the recording of the voice verification. Verification requirements depend on the voiceprint type. See step 4.

Use an audio variable to avoid repeatedly asking for the same information. For example, if identification is by account number and a variable captured the caller speaking their account number earlier in the call, use that variable to authenticate the caller.

From the Enroll ID field, select the variable used to store the enroll ID of the voiceprint.

Open the Event Handler tab.

Field	Description
No Input Event Handler	Select a task canvas. The call routes to the canvas if no input is detected from the caller after multiple attempts. The Count value tells Studio how many attempts to run before triggering the event handler. If not using an existing canvas, enter a name for a new canvas.
No Match Event Handler	Select a task canvas. The call routes to the canvas if no match is detected from the caller after multiple attempts. The Count value tells Studio how many attempts to run before triggering the event handler. If not using an existing canvas, enter a name for a new canvas.

Open the Results tab.
1. Select the variable to store the Verify Result.
2. Select the variable to store the Verify Score.
3. Select the variable to store the Verify Imposter Score.
If the variables have not been created yet, type their names in the associated fields.

The imposter score is a percentage value between 0 and 100. Higher scores indicate more likelihood the caller is an imposter.

Delete

Delete the voiceprint. It is standard practice to delete the voiceprint of an employee when that person leaves the company.

From the Name field, give the node a name.
From the Action field, select Delete.
Select the Voiceprint Type specified at enrollment.
From the Enroll ID field, select the variable used to store the enroll ID of the voice print.
Open the Results tab.
1. Select the variable to store the Delete Result .
2. Select the variable to store the Delete Message.
3. Select the variable to store the Delete Result Description.
If the variables have not been created yet, type their names in the associated fields.