Voice Biometrics

ID R&D’s products incorporate state-of-the-art technologies with voice-based solutions seamlessly. In particular, this includes speech processing, computer vision, and pattern recognition technologies. Algorithmic optimization, made by a strong R&D team, has allowed for the team to create high-caliber solutions that are agnostic when it comes to languages, noises, channel mismatches, age variances. These attributes can be incorporated into both text-dependent and text-independent solutions.

The latest version of our technology, the IDVoice™ Engine 2.0, was released in June, 2018 and incorporates three voice biometric methods including i-vector, deep neural networks (DNN) based approach and accurate p-vector approach patented by company in 2017.

Features of IDVoice™:

  • Language independent technology
  • Text-dependent, text-independent and prompted modes
  • Fast voice activity detection and speech endpoint detection in various conditions
  • Industry-leading matching algorithms for enrollment, verification and identification
  • Accurate signal quality estimation including signal-to-noise ratio and net speech length estimation
  • Configurable SDK footprint – from 2MB for mobile platforms up to hundreds of MB for server solutions
  • FIDO compatible SDK for the mobile platform

Anti-spoofing Biometrics

Spoofing refers to an attack on Automatic Speaker Verification (ASV) system, whereby a fraudster attempts to manipulate the ASV by masquerading as another enrolled person. Acknowledged attacks are speech synthesis, voice conversion and replay.

ID R&D provides technology to determine if a given audio recording contains a genuine voice or a spoof. Technologies used include: specific speech feature extraction, Gaussian mixture models, factor analysis, and deep convolutional neural networks.

Features of IDAnti-Spoof™:

  • Language independence technology
  • Very short requirements for speech length recording – from 1 sec
  • Configurable SDK footprint – from 3MB for mobile platforms
  • FIDO compatibility SDK for the mobile platform

We are happy to announce ID R&D’s Voice Antispoofing technology achieved one of the leading results from the ASVSpoof 2017 challenge (http://www.asvspoof.org/).

Behavioral Biometrics

In contrast to other biometric technologies that make use surface characteristics, behavioral biometric models use patterns relating to human activity. ID R&D’s focus is on the biometrics that can be measured by widely used devices – smartphones, PCs, tablets, and more.

The IDBehave™ SDK includes behavioral biometrics such as Keystroke Dynamics, Accelerometer, Gyroscope data, and other data points. This solution utilizes the manner and rhythm in which an individual types characters on a keyboard or keypad and synchronizes it with other sensor data. The behavioral data of a user is measured to develop a unique biometric template of the user’s typing pattern or of using a screen on the device for future authentication. Raw keystroke measurements available from almost every keyboard can be recorded to determine some keystroke statistics (including the time a key is pressed and the time between when the key is “released” and the next key is pressed). The recorded keystroke timing data, as well as other sensor data, is then processed through a unique neural algorithm, which determines a primary pattern for future comparison.

The features IDBehave™ provides are:

  • Small SDK footprint as low as 1MB
  • 6 symbols of biometric pattern or more
  • Bot detection (mobile, desktop, server platforms)
  • FIDO compatibility SDK for the mobile platform

ID R&D participation in Keystroke Biometrics Ongoing Competition (KBOC) organized by IEEE confirmed the leadership position in the Keystroke Dynamics segment (2017) (https://sites.google.com/site/btas16kboc/home).