What is Voice Authentication (VA)?

It’s not an audio recording, but a digitized set of harmonic features of your vocal tract. It proves that every human voice is unique and can form the basis for speaker identification over a phone or any other sound output. This is built like a mathematical model using commonly known algorithms such as Gaussian mixed model and Hidden Markov model. Voice authentication looks at speaker identification without bothering about content or language. Voice authentication uses basic principles of signal processing with Voice Activity Detection (VAD) and processing the same for feature extraction and comparisons.

Like any other biometrics, we have two stage process viz. “Enrollment and Authentication. Enrollment is one-time, voice enrollment and authentication, live speech through IVRS, or call center or app”. Enrollment creates a reference math score or a reference voice print, to be compared against future calls for authentication of a claimed identity.

  • High accuracy
    • Least Equal Error Rate (EER) compared to rest of
      biometrics techniques
  • Low cost implementation
    • Leverage IVR investment, if call center or self-service has been already existing with customers
  • Noninvasive
    • Just “talk” – no touch or showing face in specific angle etc.
  • Easy to integrate with existing customer touch-points apps
    • Supports both landline and mobile phones or ATMs or POS etc.
  • Can be utilized across multiple channels of interactions with customers
    • Live-agent, IVR, Web and Mobile App etc.