The Multilevel Forensic Voice Verification Model
based on Deep Learning Artificial Intelligence
Technical Background
TalkingID captures more than 4000 voice characteristics under optimum recording scenarios - features that are captured, processed and stored for future comparison.
By enriching speaker files, languages files, ambient sounds and special additional features combined with an adaptive artificial intelligence the speaker identification will be raised to a new and extremely high level of identification and verification.
Matched with an extended plausibility check, TalkingID provides insightful and unambiguous results for an identity request.
Outstanding Characteristics
All modern speaker verifications are based on the classical algorithms deriving from the early 80s. None of the existing technologies are dynamic or “deep learning”. Artificial Intelligence in modern speaking recognition (womens´s names beginning with S... or A....) is set up on top of the classical algorithms, analysing the context of ‘what’ is spoken, not ‘how’ it is spoken.
TalkingID is following a radical new way - surpassing all existing speaking recognition programs - by implementing a deep learning AI on the lowest level of phonetic analysis, learning and understanding hundreds of facts ‘why’ the phonetic of every speaker on earth is unique, by the physical appearance of the speaker’s voice. (Christian Seibert, 2021)
For the verifications it is not necessary to remember words or a sentence. TalkingID identifies the person by their voiceprint.
The human voice, a spoken sentence is the most secure biometric identification feature the human body can create. Among 8 billion humans, no voice exits two times on earth.
SmartBiometrics is introducing the radical and new software TalkingID a speaker verification that is based on artificial intelligence technology, graphic cards with 1.000 and more GPU cores enables a new quality of a deep learning forensic phonetic analysis technology, to identify a speaker 1.000 times more exact than by a fingerprint, an iris-scan or his signature.
How it works
Practically
The prerequisite is always an electronic identity of an end customer (coming from WebID, Idnow, Signicat (eID)) or directly from a merchant.
- Vocalization – Connecting voice with the user identity (one-time), voice characteristics of the person will be stored in the TalkingID database
- Verification – High quality verification process (repetitive process), comparison of the send voice profile with the stored voice characteristics
- The Result "True" of "False" will be send back to the “merchant” as result after comparing the profile with the voice imprint characteristics in the TalkingID database
How TalkingID works
Security
We are at no time in possession of a personal identity, but what we have is subject to the highest security requirements!
TalkingID Security Standard