Voice biometrics is the process of identifying a user by analyzing the characteristics of his/her speech. The great promise of this technology is that it can replace the need for typed passwords which would be a great relief for consumers stressed out from remembering more and more passwords. It would also provide an advantage to call center managers by reducing the amount of time it takes to authenticate callers at the beginning of a call.
The technology has gained considerable momentum over the last few years, partly because of the proliferation of smartphones. If you are a call center manager and think voice biometrics might be right for you, here are four things you need to know:
Let’s start by straightening out some of the confusion around terminology.
First, the more accurate term for “Voice Biometrics” is “Speaker Recognition” (see this Wikipedia entry), but that doesn’t sound as impressive, so it’s less popular.
Another term that is sometimes used is “voice password,” but that’s problematic because it leads people to believe that they will be saying their password out loud, and that makes people worry about being overheard (hence the industry is trying to move away from that term).
The truth is that a voice biometrics system does ask you to say a word or phrase out loud, but the security isn’t in the phrase itself, it’s in the way your voice sounds when you say it. If the same phrase was repeated by someone else, they wouldn’t gain access (at least in theory).
2. Why it matters
One of the biggest inefficiencies in the call center process is authenticating the user. A recent survey found that 83% of calls to enterprise contact centers required some sort of caller verification first. That same survey reported that 85% of callers felt dissatisfaction with the automated systems for handling that authentication. There are significant gains to be had in both cost reduction and caller satisfaction.
3. On-device processing is the key
The idea of using your voice as a password has been around for years. I can remember it from sci-fi movies I watched as a kid.
What makes Voice Biometrics more feasible now is that phones have become sophisticated enough to do the processing “on-device” rather than “in-call.” To understand the significance, let’s walk-through those two approaches by imagining you’re calling a bank.
With the “in-call” approach, you start by dialing the bank’s phone number. The IVR then prompts you to say a passphrase and uses that to help authenticate you. Until recently, this was the only way to conduct voice biometrics.
With the “on-device” approach, you start with a smartphone app. The app prompts you to say a passphrase which is then recorded and analyzed locally by the phone’s processor. The result of that analysis is reported to the bank. Assuming successful results, you can then have a call with an agent but skip the traditional phone-based authentication steps.
The advantage of the “on-device” approach is that audio is analyzed at optimal levels because it is recorded at the full quality of the phone’s microphone. This leads to higher accuracy and a more pleasant experience for the user, who receives immediate feedback. For the in-call approach, audio quality is limited to the quality of a standard phone call (which is a very low 8kHz). Although microphones and device processors are continuously improving, phone call audio quality remains stuck at that low level.
4. Where to learn more
One of the ways you can tell that a technology is maturing is that it has its own conference. For Voice Biometrics there is VoiceBioCon, which is organized by Opus analyst Dan Miller and has its next event in Singapore this Fall.
Nuance is the dominant provider of voice biometrics technology and they have a wealth of information on their site. They recently announced another effort in this direction called “Dragon ID”. Check out this demo video: