Voice recognition is a biometric modality that analyzes the characteristics of human voice to identify people. Widely regarded as a quick, inexpensive, contactless, frictionless, and highly secure technology, voice biometrics has already found its way into a number of everyday applications — online transactions for banks, retail, and healthcare; call centers; chatbots; AI assistants, IoT devices and even physical access control.
But it’s just getting started. The technology is expected to be even more widely adopted in the coming years driven by rapid developments in AI and neural networks that have made possible faster, more accurate, and more efficient algorithms. The speed of advances in this technology is such that some voice recognition systems have already out paced the accuracy of a 4-digit PIN.
As with other biometric modalities, voice biometrics is a “what you are” type of identification that offers distinct advantages over methods that are based on “what you know” such as passwords, or “what you have”such physical tokens. This means that it can neither be lost, forgotten, or stolen. This exponentially increases security and minimizes breaches due to phishing and other exploits.
Moreover, voice biometrics smooths the customer experience by doing away with inconvenient login procedures.
Voice biometrics is governed by the physiological aspects voice production. With at least 70 different body parts working together to produce utterances, there is no lack of characteristics and factors that make up a person’s voiceprint, which is as unique as a fingerprint. Voice recognition software can then measure these factors that make each voice print unique.
There are two ways to analyze voiceprints – active or passive. While speaking a specific phrase can speed up the voice authentication, the passive method where the person simply speaks naturally works just as well. This is so because voice recognition is not reliant on spoken phrases but on the minutia of the voice.
To verify an identity in a one-to-one matching procedure, the voice recognition system captures a new speech sample which is matched against the initial enrolled voiceprint.
In a “one-to-many” technique, a voice sample from an unknown identity is compared against multiple enrollment templates to try and find a match.
Since your voice goes with you everywhere and cannot be forgotten like a password nor stolen like a token, voice recognition is one of the most secure modalities in the market.
Yet this does not deter fraudster from attempting exploits through audio recording of an enrolled person’s voice. To thwart such attempts, developers have integrated liveness detection features into software that analyze minutia to ensure that the speaker is of flesh and blood.
A huge advantage of voice recognition is its persistence. While most modalities are “one and done” affairs where a person stays authenticated after sign-in, voice recognition can be a continuous process. Theoretically, a fraudster who coerces a person to log into a phone call via voice recognition technology can take over to carry out illegal activities. With a souped-up voice authentication system however, it is possible to do a continuous check to make sure the person on the call is still the person that was originally authenticated.
The benefit of this for banks and call centers cannot be overstated as it could save them millions.
At the moment, voice biometrics is clocking in a 0.01per cent false acceptance rate (FAR) and a 5 per cent false rejection rate (FRR) for device unlocking through biometric authentication. While this is still below the security standard for unlocking mobile device and laptop computers, the accuracy level of voice recognition technology now exceeds that of a PIN and is expected to rapidly improve.
As mentioned earlier, banking is the biggest beneficiary of this technology, with many of the giant banks such as HSBC, Barclays, and Santander already using the technology to speed up the authentication process for call canter customers.
A whopping 93% of Barclays customers have reported satisfaction for the voice recognition system which cut down transaction time by a good 20 seconds when transacting with the bank’s call centers.
Another satisfied adopter is HSBC which had noted a50% decrease in banking fraud thanks to voice recognition technology.
Slovakian bank Tatra Banks boasts of reducing transaction time per customer by staggering 66 per cent to just 27 seconds. The voice recognition system has sent efficiency soaring and has resulted in reduced demand for operators.
Voice recognition technology is also making headway in retail and e-Commerce, education, law enforcement, healthcare, human resources, and marketing.