A new audio recording of the individual saying a verification phrase is collected
The audio recording, the related user “voicekey” and UID are passed to a Voicekey server for processing
The voicekey is unwrapped and the original set of neural network weights are reconstructed
Feature data is extracted from the new audio signal and passed through the reconstructed NN for comparison
A similarity score is derived and passed out to a client system for use as part of multi-factor ruleset for determining the identity verification outcome.
Voicekey servers can be deployed on-premise, in the cloud as a public/private SAAS and combinations thereof; such that an enrolment server can be in the cloud and an unconnected verification Voicekey server can be on-premise.
Voicekey avoids the usual threats associated with hacking central databases by enabling storage of the “Voicekey” token on a mobile device.
Storing the “Voicekey” token on a portable device means Voicekey, as a product, can operate cross-channel, platform, infrastructure and organisation
Unlike other voice-biometric products, that rely on the same fixed phrase for both enrolment and verification, Voicekey decouples this dependency. Voicekey enables flexible verification options that can be used to address liveness and spoofing threats; and a better more natural user experience.
Unlike other voice biometric products that typically requirement enrolment per channel Voicekey can be utilised as a single source of identity verification in omni-channel customer environments via a single enrolment process.
Voice Verification 2.0