Overview

Introduction¶

Welcome to one of the most advanced software technology products on the market today!

ID R&D’s voice biometrics enables you to use a person’s voice as a means to verify the person’s identity with high accuracy and high speed, on mobile, web, and telephone devices. Endless possibilities exist on how to apply this capability in mobile applications, web interfaces, and call centers. To learn about those possibilities, ID R&D’s website and marketing collateral as well as the sales team provide a rich source of information. What you will learn in this documentation, on the other hand, is how to implement a solution using ID R&D’s Software Development Kit.

This documentation covers two separate products:

• IDVoice product for voice biometrics in microphone channel, accompanied by the IDLive Voice product intended for voice liveness in microphone channel.
• IDVoice CC (Call Center) product for voice biometrics in telephone channel. It also provides all kinds of auxiliary capabilities which might be useful in a call center solution, e.g. speaker diarization, age and gender estimation.

According to this categories, please use the corresponding tab in the navigation panel above.

To understand how to use the software, let’s start with a basic understanding of how the software works, as outlined in the next section.

Fundamentals of Voice Biometrics Software¶

The overall process of voice biometrics relies on the fact that each person’s voice is highly distinct due to how the human body produces sound. Therefore, by analyzing and processing samples of a person’s voice, it becomes possible to compare samples of voices to determine if the samples came from the same or different person. This is the basis for verifying identity through voice.

This is simple to say, but analyzing a voice, extracting characteristics, and creating an accurate mathematical representation that is useful for comparing voices, is a highly complex task. ID R&D’s voice biometric engine employs the latest signal processing and artificial intelligence technology to accomplish this goal. The first step is to process a voice sample and convert it into what we call a “template”. This template is a one-way transformation of the voice sample. It is not possible to re-create the original sample from the template.

ID R&D’s engine also has the ability to merge templates. For example, to create a stronger average representation of a person’s voice when using a voice passphrase that is always the same, you create three templates from three voice samples of the person saying the passphrase, and then you merge the templates with the output being yet again a new template. This new template is conceptually a stronger representation of a person’s voice because it combines multiple samples. We often call the new merged template a “voiceprint”, analogous to getting multiple samples of someone’s fingerprint.

Finally, ID R&D’s voice biometric engine also compares templates. To verify identity, you compare a new template against a known voiceprint template. If the match is high, it’s likely the same person. If the match is low, then it’s likely not the same person.

This SDK drastically reduces complexity for the developer and provides these capabilities to create, merge, and compare templates through API method calls. These three functions are the core of what you need to implement in every voice biometric application. The SDK also provides configuration options and a variety of convenience functions to make your life easier and deliver a robust user experience.

However, performing a match or no match does not make a complete biometric system. In many situations, a bad actor may try to thwart the voice biometric system. To detect attempts to attack the voice biometric system, ID R&D also produces market-leading technology for voice liveness check. This technology detects whether the voice sample is a live voice from a real human and not a recording or a synthetic or converted voice. You access voice liveness check with a single API call. There are some limitations to be aware of in telephone-based applications, but overall, voice liveness check is an essential element of nearly all voice biometric deployments.

Our technologies¶

Our technologies:

• Kill Passwords
• Minimize / Eliminate authentication UX (“Friend Experience”)
• Close Security Gaps (Liveness)
• Bring biometrics to Conversational Interface

ID R&D combines best in class biometric technology (Voice, Behavioral and Face*) with Liveness Detection for continuous authentication and a frictionless UX:

• VOICE - Our industry leading AI-based voice biometric engine developed by NIST winning team delivers high accuracy, low footprint and robustness to noise
• BEHAVIORAL - Everyone types / interacts with their devices uniquely (like handwriting). ID R&D developed machine learning algorithms to identify these individual patterns.
• LIVENESS - ID R&D is a leader in voice and face liveness with a unique AI-based approach that can distinguish a human voice / face from a digital representation (recording, photo)