ID R&D used its presence at the Voice Summit 2019 in Newark, NJ, to introduce the latest version of its flagship voice biometric platform, IDVoiceTM version 2.7. The release steps up the integration of elements of artificial intelligence and voice processing technologies both to shorten the time it takes to enroll a voiceprint and, probably more importantly, to speed up “biometric matching speed” by a factor of 10. That means that it takes seconds to authenticate the identity of an individual speaker. The result is the promised “zero effort” speaker authentication – which I think of as Speaker ID-lite – that can smooth the way to Voice Commerce through a range of end-points, from smart speakers to Interactive Voice Response (IVR) systems and contact centers.

ID R&D positions IDVoice v2.7 as a platform for enterprises to implement voice biometrics across multiple communications channels. It joins NICE Systems and Auraya (Amorvox) in emphasizing its ability to use a single, enrolled voiceprint to be used in contact centers, native mobile applications or web applications. Thus it saves enterprises and their customers from such time-consuming and off-putting processes. The new release makes it easy for developers to account for characteristics of different channels automatically without requiring a separate enrollment. They can set a parameter in the IDVoice API to make ID R&D’s biometric engine aware of the source of a voice sample. The engine then, automatically recalibrates the probability scoring and acceptance thresholds, taking into account whether the enrollment source is different from the verification source.

From its inception IDVoice was designed architecturally to be independent of channel, language and age variance, as well as highly resilient to environmental noise. It runs on iOS, Android, Linux, and Windows, in order to accommodate a variety of mobile devices, servers, private clouds, and embedded IoT architectures. In the spirit of supporting the democratization of speech, IDVoice v2.7 includes wrappers in Python and Java, designed to appeal to enterprise developers, and IDVoice v2.7 also ships as a Docker image. The Docker image enables initial testing with minimal integration and automation for production deployments, an ideal scenario for creating a voice authentication server for the enterprise.

Other enhancements to IDVoice embedded in V2.7 serve as a “tell” of where secure and personalized #VoiceFirst services are headed. The new release adds improvements in the platform’s Signal-to-Noise Calculator and Voice Activity Detector which, when combined with the 10x speed improvement for authentication makes personalized, secure “wake up words” a definite reality. The speed bumps that have slowed the roll our of #VoiceCommerce are rapidly being eliminated; accelerated adoption to follow.

Categories: Intelligent Authentication