CEO of UCIC — helping hardware products come alive with voice interaction. Proofs of concept, prototypes, and tools for integration of voice services.

Aug 7, 2016

Voice Biometrics

Voice biometrics is one of those technologies that has been around for decades but is going to see its renaissance over the next two years.

Just before Robert Redford was making indecent proposals, he was breaking into super secure facilities through crude voice biometric hacking…

This popularized the vulnerability of voice biometrics. Someone might have made money by shorting companies in this space before Sneakers came out in 1992. Voice biometrics were flung to the far corners of interactive voice response (IVR) systems of banking or government customer services. The enrolment process was clunky, the system requirements were high, and the reliability ultimately was low.

However, today the renaissance is coming through a few new technologies:

Easy and quick user enrolment. Fixed trigger with user enrolment (in the parlance of companies like Sensory) are simple to implement and not only identify the user but also enhance voice triggering.

Non-fixed detection. Speaker identification can now happen without the need for a “my voice is my password” type phrase. To get around the scenario in Sneakers, voice biometric systems will ask for the user to speak back a random phrase. Sorry, Robert, you’re stuck in between two panes of bullet proof glass…

User liveness detection. This is a bit of black art — algorithms can be implemented to determine that the voice is coming from a live person rather than through a recording.

Multi-factor authentication. Of course, just being able to do things like detect the presence of the user’s cellphone, have them speak a pre-known code, or use some other additional means, voice adds to the verification.

Because ambient voice interaction is slowly marching towards the early majority adoption curve, there’s going to be a growing need for voice biometrics. Especially scenarios where a financial transaction is involved, such as purchasing on the Echo. There are dozens of companies that have been working in the background in this field that we’ll likely see take off in the coming years.