ArmorVox™2015 – A Focus on Passive VerificationApril 23, 2015
In January 2012 Auraya Systems released ArmorVox7; the first voice biometric product to “fused” Text-Dependent and Text-Independent technology into a single software license and enabling partners to implement both Active and Passive voice biometric solutions (or a combination of the two) using a single software license. Since that time, others have followed suite validating Auraya’s original innovative approach.
With the release of ArmorVox™2015, Auraya has now brought together over 10 years of R&D and market research to introduce the world first Universal Speaker Recognition Engine, that carries forward the fused technology approach to an ultimate conclusion – a software license embedding all the technologies, tools and features developer need to implement any voice biometric solution, be it for Government, banking, financial services, healthcare, IT services, smartphones, law enforcement, national security and military.
New to ArmorVox™2015 is an extended front-end that improved accuracy and new noise cancellation technology that improves robustness in noisy conditions, especially in mobile phone channels. New acoustic filtering techniques extend the accuracy further; especially for cross-channel solutions where speakers enrol in one channel (say mobile phone channel) and verify in another (say landline).
But it is Auraya’s unique tuners that contribute most to performance improvement. Unlike our competitors that ship generic models with their product; Auraya ships the tuners embedded in the product to optimise the models to the accent, language and communications environment of the population of enrolled users. So irrespective of the applications and the linguistic environment, ArmorVox will always optimise accuracy for the users enrolled in the system; which, after all, is what you want. It is no use having a system optimised for, say English, if all the speakers are Chinese.
The upshot of all these new capabilities is a two to four fold improvement is voice biometric accuracy and robustness.
Up until now the Auraya’s tuners have been available for the text-dependent only. Whilst, the text-independent and text-prompted technologies could be tuned, this was a manual process. But with the release of ArmorVox™2015; Auraya’s tuning technology, which has proven so powerful in text-dependent deployments, is now available in all other modes too. So the same level of accuracy, robustness and reliability seen in Auraya’s text-dependent deployments is now available in text-independent solutions, including passive enrolment /verification and speaker identification systems used in fraud detection and tracking.
This new capability allows end-users such as banks, Governments and financial services, to utilise their large stock of voice logs captured from previous customer calls to bootstrap their voice biometric deployment. These recording can be exported from their voice logging systems to ArmorVox™2015 to create a readymade voiceprint database from which to authenticate customers without need for explicit enrolment. This makes enrolment and verification “frictionless” for the customer; thereby enhancing the customer experience. And even if the voice logs are mono (that is both customer and call centre agent in the same channel), Auraya speaker splitting technology can separate the speakers and create voiceprints for both sides of the conversation.
The real benefit of ArmorVox™2015 is that it unifies all voice biometric modes in a single system. This way, both text-dependent voiceprints, used for example in smartphone apps, or old-school telephone banking solutions, can be fused with text-independent voiceprints for passive enrolment and verification in the call centres. No longer do systems integrators and applications developers have to choose a technology before building their applications. Instead they can concentrate on building the application that best meets their customers’ business requirements and utilise the biometric feature to enable their solution.
ArmorVox™2015 brings together all the benefits of voice biometric packaged into a single unified software system.