Why VoiceVault?
Using the uniqueness of the voices of your customers, your business can quickly, easily and cost effectively use voice biometrics to verify a claimed identity.
High accuracy from short utterances
Our voice biometric technology is able to achieve high accuracy, speed, quality, predictability, and testability in the verification of a person’s identity using a small number of short utterances. For the amount of text that we use (typically of 5s duration) the accuracy of our technology is unparalleled.
Our accuracy is also largely independent of:
- The level and type of background / or environmental noise
- The length of the utterance
- The variability of the channel between enrolment and verification (or verification attempts)
- The ‘state’ of the individual (health, mood, narcotic intake etc.)
Independent of channel effects
In our technology the words or phrases needed to produce a biometric voice model are system-defined and specific. In such a system, the user is prompted to say specific phrases that are derived from a small vocabulary with known phonetic content. This ensures that the voice model reflects as much of the user’s voice biometrics as possible which in turn has a significant impact on system accuracy.
As we prompt the user for the speech that we want them to say, we can check that they have indeed said what we asked for. As a result of this, we can filter out extraneous noises in the channel or from the environment that the user is in. In practice this means that a user is able to enrol on one channel and verify on another with no impact on performance.
Short enrollment and verification
The enrollment process varies depending on the mode of operation of the voice biometric system, but will normally involve the user repeating a number of prompted phrases such as “0 5 7 9”. Typically, four such phrases (about 15s of speech) are sufficient to achieve very high accuracy levels – and of course a good user experience.
In the verification process, the user is asked to provide their claimed identity, typically an account or card number, before being asked to provide a single short phrase – about 5s of speech.
Replay attack detection
Biometric systems, as with all security systems, are vulnerable to fraudster and impostor attack. In a voice biometric system these attacks are generally based on the use of one or more recordings of a genuine user.
VoiceVault technology is able to detect replay attacks in real time.
Language optimization
In our technology, language-specific word choices are made that best expose the vocal tract characteristics for that language.
The use of a language pack enables a system to be optimized for a particular country or language. It is also inherent in the design of the language packs that they continue to refine the accuracy of the system as more and more users of that language enroll and verify themselves using their speech.
