PerSay allows a developer to integrate their VocalPassword product either with SOAP web service calls or with VoiceXML <data> tags. For the examples here, we’ll use <data> tags. When enrolling or verifying a caller, you simply <record> the caller and use a <data> element to post the recording and caller information to the PerSay server.

The steps to using PerSay to authenticate callers are:

• Sign up for a PerSay account
• Establish a PerSay session in your VoiceXML
• Recording the caller’s passphrase several times
• Enroll the caller by submitting the recordings to PerSay
• To authenticate, record the caller speaking their passphrase
• Submit the recording to PerSay
• Check the PerSay result

To start using PerSay Vocal Password contact PerSay for a trial account.

An application using PerSay starts out by creating a session for the caller. It can then enroll, re-enroll, delete previous enrollments, check if a user has already been enrolled, and verify a user. For the purposes of this simple application sample, we’ll assume that a caller is not previously enrolled and simply enroll them. Then we’ll verify their authenticity. In a real application, you’d want to check to see if a caller was already enrolled, then either remove their existing enrollment or skip enrollment, but such detailed call flows are beyond the scope of this guide.

The sample application will greet the caller with a VoiceXML prompt, then load a locally-hosted subdialog that will generate the session. The session ID is stored in a global variable so it’s accessible to all the other subdialogs.

Once a session has been generated, the application will attempt enrollment by launching a locally hosted enrollment subdialog. The subdialog asks the caller to repeat their password (0123456789) three times, records the audio, and uses a <data> element to send the recording to the PerSay server. The subdialog checks for error conditions and plays error messages to callers as appropriate. Once enrollment is complete, a result flag is set to either “trained” or “error” and call flow returns to the main VoiceXML document.

After enrollment, the application will launch another local subdialog for verification. This subdialog will ask the caller to say their password and then records the audio, submitting it as a post request through a <data> element to the PerSay server. Again, variables are set indicating the result of the verification and common errors like silence are handled by the subdialog.

Call flow once again is returned to the main VoiceXML document, where you can read the results and make decisions about what to do. In our sample app, we thank the caller for a successful verification or admonish the caller of their error and then hang up.

You can get the sample VoiceXML application from Voxeo’s GitHub account.